 Hello everyone welcome to the course of business forecasting. In today's session we will discuss a new concept that is called expert judgment in time series. So far whatever the models we have discussed under the context of time series analysis it was relied on or based on the historical data or say past data. Today in this session we are going to discuss the experts opinion or the human judgment of contextual information for the future event and how that can be integrated with time series data how you can make a revised forecast of your time series forecasting. So let us see how it does it work and how it can be designed. So far if you go back to the just basic information I have mentioned here about the summary of previous time series information and the new context that we are going to discuss today. In basic mathematical forecasting or say time series forecasting are deterministic where your data are historical and based on that you know the pattern of the data and you are making a forecast by analyzing the behavior of the data we are making the forecast for the future. So this part you can see on that particular information. Now look at the second point where some unpredictable events I will discuss more detail about that what kind of futuristic event future event or unpredictable event that can occur in that case your past data or historical data may not be suitable to make the accurate forecast. Based on the historical data or say time series information or modeling you can make a better forecast with minimizing your RMS or whatever or maximizing R square that we learned. But in this particular example or particular setting what we are going to discuss is that some future or unpredictable events may be so significant that your past database forecasting may become absolute. For example let me give one example. Suppose now we are seeing the Israel and Palestine war last year during this time period we do not had you know this particular war in Middle East but now this year it happened what happened because of this event new event the crude oil price are getting fluctuated it is going up. So if you take the past couple of year data of that particular time period your data may be smooth or you know you might take a based on the basic trend of the data you can make a forecast. But now because of this particular event unpredictable event what happened you know your crude oil price has gone up and that is only experts can assess it your past data cannot assess it. So if you expert feel that this going to impact this particular war going to impact for say 3 months or 4 months of you know crude oil price and which will have a big impact to the GDP or you know different you know use of crude oil requirement. So in that case what happens many companies forecast will have a impact on that. So in that case even crude oil price if you want to make a forecast so you have to incorporate this particular event into your data modeling for the future couple of months say. So this particular you know unpredictable event how you can integrate the information how you can capture and how can you integrate into your previous time series data modeling and how can you make a revised forecast of that that we are going to discuss today. So now as I mentioned so here we integrate the basic mathematical model of the time series model along with that domain knowledge experts opinion right. So this two part we are going to integrate together. So what are the events that could happen the special event some suppose you know suppose Amazon is giving big billion days sudden you know discount offer etc. So that will have a big spike in your you know say sales say or say you know government is conducting some event or say you know tech face or so many other information can come which can have a big spike in sale or you know lack of demand or some you know I will show you more information about this with examples so that you know the curiosity remain there with you then strikes can hamper the sales or so you know can hamper the productivity or you know you do not expect based on your past data. Similarly promotion etc the other couple of events that which can impact your decision making you know in banking sector different type of you know catastrophic information or say you know fraud case can have a impact in your share price. So many many examples I can give which your past data cannot capture and your future event can also be a value added to your revised forecast process. So therefore the idea here presented is that you know the structuring the impact and explicit knowledge in order of the experts in order to easily and fully integrate it into the time series forecast. So this integration part we are going to discuss here I have listed couple of examples to get better insights about the concept of intrication of human judgment or experts opinion in time series forecasting. You can see couple of examples here say you know Russia-Ukraine war the impact of sudden event of Russia-Ukraine war or say you know Israel Hamas war what are the impacts it was sudden event and what are the impact the sudden spike of crude oil price and the impact in economy of different countries. Energy crisis due to Ukraine war the energy crisis in entire Europe it could be a jump factor I will discuss more detail in the foot coming slides you will get to know think about Israel Palestine war it might last for three months six months and after that it will come down to the normal price. So these are the sudden event and it will not last for longer period maybe couple of months couple of season it will last after that it will come back to normal situation. So these are called sudden event earlier it was not there in the previous years before Ukraine war or say before Israel war this event were not there in the previous years of data but in future it has come. So today it is an event contextual event next year this event the impact of these events will become a out layer. So if we can consider the out layers in your time series data and if you clean the data why can you consider the future event such type of events in your time series forecasting because you know that is going to happen expert know but that data does not know. So you want to integrate that some aspects of that through contextual information through a systematic process you want to revise the time series forecast. Here you can see few more examples supreme code verdict on reliance acquiring future group few years ago it happened I know and it had a huge impact in future group share price. Sometimes you know Amazon wanted to acquire the future group reliance entered into the picture then it went to the supreme code in Singapore code and then the entire future groups situation the assets are on turmoil what happened future group share price crashed and it is a permanent it has gone down it is not recovering now. So it is an impact sudden impact like you know future group share price were there but suddenly it started falling. So it is a impact and it will remain for longer period you might say longer period impact can also be considered under human judgment in time series. Yes there are different aspects which I will discuss in the forthcoming slides many type of contextual information can be captured which is a part of integration of human judgment in time series forecasting session. Here you can see few more example like you know the CEO of Chanda Cocher you all right what happened because of our husband some malpractice what happened ICIC bank share price crashed or market share they lost the market share even you think about the example of say you know the event of World Cup in Amidabad city what happened these three four months there is a huge page about hotels booking and the logistics part etc many companies got the advantage of that but after three four months that once the event got over before the event there is a big demand and after that it become to the normal mode so this is called you know contextual information. So that expert nobody or past data does not know we want to integrate that. Similarly you can think about the example of you know shifting of IPL matches from India to UAE during corona times when corona came suddenly you know the IPL man has to be postponed then after few months I think they shifted the remaining matches to the UAE so it had an impact of logistics and the you know the entire IPL event and the companies who are who were associated with that they lost the opportunity in India think about the example of say you know the inauguration of Ram temple you know what happened many hotels are coming up you know and maybe couple of months the crates will be there and after that it will come back to the normal mode few companies few hotel will start you know making their new businesses over there. So these are the new event or contextual event it might last for few three months six months but after that it will be normal and couple of companies will last as a permanent aspects of that. So these are the contextual event in the past it was not there in your time series data and if you make a similar forecast with your past data your forecast will be like this maybe but now because of this event it might go up it might be like you know step function kind of thing also depending on the pattern of the business and the requirement of that so after that couple of time period it might go come back to the normal mode also. So depending on the business model you will have a contextual information and the impact you need to integrate it with your time series data and then you should make a forecast because you know these events are going to happen and that will have a big impact in future of the sales and the marketing plan. So why can't you integrate with the before the event starts and that partial information of that we are going to integrate through contextual information through quantitative aspects of analysis with your time series data. How to integrate these informations and revise your forecast that we are going to discuss here. Here you can see few more example like you know sudden weather change impact and the predictions of the upcoming week sometimes it happens you know in western country when sudden you know big cold comes it impacts the you know business models of the society or the you know different type of social science projects. Look at the example the launch of geo and its consequences and impact on the telecom companies many companies actually bankrupted they actually stopped their business almost you know no one is buying their products because of new entry of geo. So this is what the impact it is a permanent impact you can see before geo came it was not there that this couple of companies were performing very well but once the geo enter into the market in the competition what happened it captured the market share suddenly people shifted to geo. So it is an impact so the experts know it think about recent trend of government decision say you know railway restructuring process most of the railway companies are performing very well you know think about renewable energy sector government is coming up with different PLI schemes it will have a big impact but in the past it was not there in different sector also you can see whenever government is taking some initiatives it will have a big impact and the experts on that field know that prime minister is going to announce some scheme or finance minister is going to announce some scheme or the concerned ministry and that will have a impact in the share price or say you know in the business models of that particular sector and you can play on that in say in trading of stock price or say you know in business models or investment planning company can think of that and they can think about in a product line extension or new investment planning. So these are the contextual information in the past it was not there but if you know that this is going to happen it will have a impact in your business why can't you modify your revised forecast or your time series data and make a better planning of your decision making it not only about forecasting it is also a statistical decision making also. So look at this example Navy Mumbai airport construction or say MTA HL the hotel hotel setu it will have a big impact in the real estate and the business models of Navy Mumbai area earlier it was not there now these two big projects are coming up it will have a drastically change game changer event you can say in that particular area it will have a impact and it might last for longer period so it's a contextual event earlier in the past data it was not there but now you have it so but it has not come yet. So you are trying to integrate that impact in your previous data and you want to modify your forecast this is called human judgment in time series forecasting let's see with technical illustration of that now. So now what are the factors remember it is not easy to integrate the entire information like future event you don't know what would be the impact and how much you should consider so that is very questionable part. It might not happen also suppose you know you might think that something is going to happen expert are saying that this is going to happen be ready with that and you modify your forecast but suppose in practice it does not happen. So statistical people sometimes feel that no no no we should not integrate the human judgment the future event we should rely only on the past data and based on your time series regression of the you know machine learning process we can do it. But this is also part of your human judgment right this is also very important in AI process we should do that that means what is going to happen if we can capture that well in advance if you can modify your past data forecast perhaps it is a very good impact in your decision making there might be error but if you can put effort on that then it could have a big decision making your statistical decision to the company. So that how to do that the future event information may be very small as compared to the past data but the impact is too significant to neglect here you can see that the occurrence of coming events can be recognized in advance although it is not easy to judge their impact precisely how much will be the impact you can even imagine that that you know Navium by airport and the say you know MTHL bridge will be a big game changer to the entire area of Bombay but how much will be the impact how much you should consider that is also very challengeable and the experts opinion only a source of integrating that information or collecting the information so you have to rely on experts. So when you rely on experts how much percentage of that experts opinion or the trust or the faith you have or the experience they have that you can integrate that we will discuss through some quantitative approach of like conversion now the point here is that so far whatever I have discussed the future event different you know strike promotion whether political or say you know or many example right I have given corona many examples I have given strike many thing can come and which could be you know only experts say stock price insight trading information anything so this are the future event and that you want to integrate with your past data so effectively you have the past data and you want to make forecast for the future but what happened you know you are here say you are standing here now now future event can go you do not know but you are assuming that future event will have a impact your forecast it should not be like this you are you are predicting it right so all these events how to capture how to you know quantify them and to integrate with your model that is the question now the context the background the motivation we understood but how to execute it so there are four type of human fact judgment factors that you can define and throw which the future events can be categorized and you can integrate through this channel of four factors what are the four factors this has been developed by you know this Neufel et al and his team so this reference also I have mentioned here you can go through detail of it so that information that paper reference we are you know exploring in this particular course of human judgment in time series I like that particular work and so thought of sharing this information as a part of your business forecasting model so now let us see the first factor is transient factor I will explain all four one by one the second factor is quantum jump factor the third is you know transport impact factor transfer and then train change factor so this four factor can be defined through which all the future event can be put under this four factor so what are the four factors and these four factors in general there are many way you can capture but when it comes to the business context generally it could be declassified into three different category like the market conditions can have an impact through these four factors or say client issue can come or say contractor level like upper stream supply chain level can also have an impact on your decision making the supplier contractor you know all these things can also have an impact on either client and the market so these three you know category you can define from where you can get this particular event based issues but all of the information that I have said can be categorized into this four category transient quantum jump transfer and train change suppose you have captured the data sets the future event through these four factors and then you would like to integrate how to do the methodology the integration process first step is that you develop your mathematical forecast or say time series forecast as it is whatever you have learned that the techniques that we have followed any method for the data pattern once you understand the you know the data behavior etc and you select the model and you fit your forecast so basic time series or the regression what about the business forecasting models we have been learning so that you can use so you first you make your forecast done forecast are done ready by removing missing data etc and then clear your data understand the train and fit the model done then you might say sir it is forecast done no this is stage one then in step 2 what do you do you identify what are the contextual information based on that you have already made your forecast right you have made one forecast say now what you are doing that you understand or capture based on the experts opinion based on the experts opinion you identify what are the event that going to happen in future next one year or six months in a particular time period or immediate whatever the problem statement context you have you understand the experts opinion you collate it and then through this four factor that I have talked about through this four fact electronic transport you know train change etc you collect the data through this particular factors as a part of event and then you give the weight of that impact of that particular event how much impact will be there through the weight multiplication which I will show in next slide and then once you collect that that part is ready as a part of how much impact you are collecting in a structure manner so that part you do as a second stage decision making and then in third stage you integrate your time series and the human judgment event with your impact right impact right so these two integrations addition of this part will give you your final forecast here I have shown you the steps so this step one is nothing but the step one is nothing but your basic time series basic time series as it is stage two the step two you can see step two are nothing but the capturing the factors the event that going to happen through experts and then you are collecting them through this four factor of human judgment right so once you collect that data you split it based on the time and we segregate all this data and add this impact this impact you add with your final forecast so this plus this one will be your adjusted forecast and then this is your final forecast this adjustment has to be added into your model so that you can make a revised forecast so this is called step three look at here so this is the impact time wise time dependent impact look at time dependent impact impact over period of time so different time will be there and the how much impact you are getting for these four factors you add them you say aggregate them and for overall all the time period how much impact you got and then you add that that you include to your final forecast time time dependent in each time period how much forecast for example you have a time here like said time you are here some impact will come from say you know transient factor some impact may come from say you know train change factor suppose and say transport factor so four factors are there so from this four factor all four may not come here suppose one might be there two might be there so you add all this factors impact and then that you add to your time series forecast model and then you revised it so this it is not difficult but a structure way we have defined it so let us see how this process works how much impact you want to consider of a experts opinion for the event you do not know like you know that mthl bridge or say nevy boom by airport will have a big impact of the real estate price in the nevy boom by area suppose the example that i am giving or say you know noida airport will have a big impact in getter noida region so but how much or say delhi bombay expresso will have a big impact in say couple of cities nearby that particular expressway so how much will be the impact on that particular event or the you know project say so that future event you are going to capture say or say you know say the example that i was giving the inauguration of ajeeta temple will the ajeeta city will have a big boost on that you know the market on that particular region so how we will capture that how much the experts might say that it will have a big impact but how much you want to capture based on the experts opinion the reliability of the experts and the the truthness of the experts will have to define your revised forecast or that portion of that you have to consider suppose the expert says that the maximum impact could be delta max suppose so it will have a say you know thousands extra amount of room bookings say so if experts say the maximum it can go to thousands but how much you should take based on the experts opinion the group decision making you should take probability or you can say the weightage of that based on the reliability the past experience and the confidence of that particular you know consulting firm do that so you can see that particular analysis through you know conditional probability or say you know Bayesian analysis perhaps you all know right so that way you can guess how much reliability the experts have or how much accuracy they predict based on that you can take percentage of that if you are fully confident about the experts opinion that you have a good trust so you can take the because company will have to modify their forecast right the business planning right so generally 80 percent above considered as a highest impact like your weightage will be high above your maximum impacts at delta max and then out of that how much you want to consider if you are fully satisfied you can take full if you are partially you can reduce your weightages so less could be less than 20 percent also and then you know medium could be 40 to 60 percent like this way highest could be above 80 percent so this way you can classify the impact of the experts opinion maximum impact you can define based on the group decision making of couple of experts opinion and then once you get the average based on that how much you want to consider based on your confidence or how much you want to modify your your past forecast right that mathematical forecast of the time series forecasting that that you need to consider so that is called you know say I say where I is the final impact where is this is the maximum impact because of that event can happen whether it is a transfer factor train change factor and part of that how much weightage you want to give or how much percentage of that you want to consider in your decision making so this is your impact part now let us see the four part the impact part you understood how you can add all of them and you can add it your future model right the revised model here you can say revised model how can we add it so now the point here is that what are the four factor that I was talking about so first is the transient factor remember this is very important four factors examples so the transient factor or nothing but suddenly that event occurs and it lasts for couple of period and then it goes like let us say welcome match in amedabad city there is a craze of say one month two month of hotel booking restrooms and all these things to visit that particular city so there will be big spike of say businesses related to that particular sports so there will be spike sale so that particular couple of months is called your time frame and that during time frame you will have a big spike after that once the event got over you will see there will be no impact of that particular event so that spike part you have to integrate you know that that could happen and you are trying to integrate in your time series data so that future event you are capturing throat transient factor strike sales promotion etc could have an impact on that there are many examples which you can define once the event got over it will not have a impact to your time series data forecasting or the decision making but during that event there will be a big impact maybe up maybe down in your data process or the forecasting process so you should integrate that you should capture that throat transient factor but the question is how much of that you should consider so that I have explained percentage of say delta max is the maximum impact out of that you take percentage of say w into delta percentage of that opinion the expert's opinion based on the reliability expert's opinion and the experience of the consulting team you take that now transient factor we understood transport factor transport factor are nothing but about the risk kind of thing or you grab the opportunity suppose somebody is giving say big discount some retailer or say you know some manufacturer are giving big discount so retailer would like to adopt this opportunity so they have a regular ordering process to the manufacturer as per their contract but now because of the discount offer suppose a manufacturer would like to reduce the inventory so they can get an advantage of the discount offer so this is called transfer the next month they might not buy that but because of the discount they are going to buy well in advance so this type of policy you know discount offer or say to some extent you know sometimes say bonus issue given by companies so many examples where you can see people or the clients get the opportunity of that and in that case what happened your sales or the demand or sometimes that gets shipped so this is called transfer factor you can see the demand whereas it is regular but it has shifted here now it has gone down it has been adjusted here so or like you know supposedly your demand planning was like that you know overall and suddenly because of this discount it will be like this but during this extra sales or the opportunity the discount opportunity the retailer might buy more order here rather than this time so since they have bought here naturally with discount price naturally next two months they may not buy so it is called transfer factor there are many such examples which you can think on that particular category another factor is called you know quantum jump factor suppose you know suddenly some magnitude change in the demand say since it is a business problem I am giving this example but you can relate with any practical social example also quantum jump suddenly suppose your computers close their stores so enter demand will shift to your particular store so what happened that enter demand shifted to here or you have come up with a new product you will see the big spike in your demand say or and it will remain steady say so once it is there but once because of this event you will be see the big demand in that particular product so it is a big spike of and then it will remain there it will not fall down again so then if it is fall down once the event get over that is called the transient factor but it will remain stay so that means it is a quantum jump it may be up it may be down also so you do not know what will happen in future but that you are trying to capture through experts opinion if you think that it will be up so you consider the additional information of your time series data forecast if the experts feel that no because of that event it will be down so you consider that down part is also good right if it is down if your sales will be down if your expert know that the sales is going to down and if you do not rely on that if you rely only on past data and if you make your procurement planning production planning etc you might say big amount of cash engagement but if you rely on experts opinion experts saying that sir this is going to happen and say you know you should be ready with that therefore you can integrate that contextual information and you can modify your data whether job or down both situation are good actually because you are revising your forecast now last one is a trend change factor trend change means suddenly over a period of time you can see the price variation over a period of time that will have a sequential changes suppose some company is doing a very you know bringing innovative product in every quarter say so you will see the demand of that product will be up so over a period of look at the trend is changing up trend or down trend suddenly you had a trend but suddenly it will be like this or suddenly it would be like this so in that case that factors you have to capture into your model like you know let me give you another example suppose you know during the like in Ukraine war and the post corona post corona and the Ukraine war what happened RBI in India or say you know the Fed in US they started increasing the rate most of the experts on that market the stock market or say you know in the financial sector are predicting that both RBI in India and the Fed in US also probably will cut the rate in India the RBI probably will reduce the reporate so therefore experts are predicting it and that will have an impact in your say you know stock market or say share price etcetera who are sensible to rate cut say so this future event if you can capture but when RBI will reduce the reporate it does not mean that they will reduce in a one at a go so they will reduce slowly slowly so the more they over a period of time if they reduce the reporate one by say 0.5 0.25 0.25 like this way if they reduce the reporates over a period of one year two year you will see the spike in share prices related to say NBFC banking sector you know metal or say you know IT industry there will be impact on that so this is called the trend change factor so there are many examples which you can integrate through this factor of trend change so these are the four factors here I have given a summary of this four examples also you can see one example another example say you know quantum jump factor you can see impact of non-repetitive event is permanent and here you can see the addition of a new customer say trend change factor here you can see the factor modified the demand trend you can say the increase in price could have a trend change factor price increase could reduce your demand say and then say slowly slowly slowly you are increasing price and slowly your demand is getting down transient factor you can see the strike sudden spike and then go down or sudden down and normal come back to the normal and the transport factor you can see then a discount sales offer for a limited period so the amount will shift to the previous period and the sales will be adjusted accordingly so these are the four factor that I have mentioned and there is another example I have given and here is the impact potential impact but how much weightage of that you want to consider so multiplication of that is your final weightage w1 into delta 1 that you should include to your time series data set so this way the four factor you can integrate to your any example of time series with future event in the future context right now you can apply it in any particular example say any practical case studies here I am going to show you one case application based on this particular theory the concept that we have learnt through this example you will be able to understand the better about this particular application of expert judgment in time series in this particular example here we have taken the demand forecasting for a plastic bag manufacturer these data also you have taken from that particular paper that I have mentioned and that example I will be the case example I will be illustrating here right so the reference I mentioned so we can take that same data and we can explain it right so suppose the plastic bag manufacturer company in Europe and the researchers would like to study the impact of experts opinion or the image as the future event contextual information into their data and their forecasting process so how they have done it earlier they had the three year data of the company plastic bag sales data they had right now they would like to capture the future events if any further particular company throw this four factor one is the plastic price of the products and the special offer if the company is planning to give and say plastic bag improvement suppose you know they actually wanted to launch some biodegradable product which a new environment friendly bag so that opportunity can have a spike of the new clients for that particular product and another is a client situation so this four factor they have captured now this four factor will be related to the four human judgment factor that I have talked about what are them transfer factor jump factor transient factor and trend change factor right this four example will be linked with this four factors say suppose here you can see the transfer factor what is the transfer factor it is actually you know this special offer the special offer or nothing but the transfer factor because of the discount out of four five clients large clients of the company so they found that you know two clients to some extent will have like you know will accept this you know like you know discount offer special offer so that amount and the impact we have captured here similarly jump factor you can see because of this you know this particular new bag biodegradable degradable or say you know environment friendly bag they found that one new client will come they are expecting that one client might be in touch with them and they are expecting that the client will buy such amount it is in in the thousand multiplication of that data so you can see it is not 200 200 multiply thousand in terms of number of bags say and the weightage could be around 80 to 100 percent sure that this fellow will buy this much of amount and the new bag will be sold out through this particular client there may be other clients also regular clients can also buy but they are sure that the new client has given a order or given a has shown interest to buy this new product that that company is planning to launch there is another transient factor what are the what are the transient factor it was like you know client situation what was that suddenly that one client i think client number three say for example had may had informed that that in a particular post summer this a case was from Europe so perhaps in the September month that one client has shown has informed that company manufacturing company that post summer they are going to renovate their store or they are you know premises so they won't be able to run that one month of their business so they have made a request that can i stop buying that particular product for that month because there will be no sale and the company feel that it's a good client genuine client so we should accept the request of the client and we will be assuming that since the client will not buy that particular product who are regular for our business as a regular client so we should accept and the request and we are assuming that since this client will not buy that particular product because of the renovation of the store so there will be no sale of that product from that particular client for that particular month next month onwards it will be coming back to the normal mode because they will open the store again so this is called the you know transient factor suddenly there will be a fall for that particular month and then down because the client has mentioned that they will not buy that product because of store innovation process and the another is that the company because of the raw material price company thought that there should be you know last several months they have not you know increased the price of the product now they would like to increase the price of the product but you know one not at a one at a go there will be loss in the demand so they want to increase the price slowly slowly for the next one year and then there will be impact of sales also just this is a sample example you can understand sample case study so they will reduce the they will increase the price and because of that they will change percentage of change in their you know sales so they are assuming that on an average say you know two percent changes will be there so the company feel that there will be you know sales down once you increase the price in a monthly basis so there will be sales down in sales also so that information also been captured so look at this the summary sheet so the four factor that I have mentioned plastic back price increment will have a will have a impact in your change change change factor then special offer that I mentioned which is nothing but transfer factor discount offer so this particular two three client that I have mentioned in the previous slides will buy that accept that offer and they will buy in this month of January so therefore they will not buy February month so it will shift now jump factor you can see the plastic back improvement because of that couple of new clients they have found and because of that they will see that you know this particular time period there will be extra sales because of new client or because of new product testing right they are planning to launch in the market well in advance they are predicting it before the next year starts the client situation one client mentioned that you know that he will you know go for innovation process of their store so he will not buy the product he will not run the store for that particular month so there will be no sale on that particular month so that is been added also the impact has been captured with say multiplication of the percentage of delta and the weight and the total impact say so this is the overall data say now what you are going to do this you in each time period in each time period you are capturing the impact right the four impact if at all it is there then fine then you add them this addition is nothing but your total impact for the time period once say now for time period to calculate the all four impact all four impact you add them right you add them just you just add all of them you will get like for this month you can see there is a you know sales down because of your plastic price increment so this is called trend change factor trend you are seeing that every month that is a less sales because we are increasing the price the company policy and suppose you know on that particular month you could see that there will be no transfer factor there will be no you know jump factor but there will be a you know transient factor because of down because of you know closing of a store or say you know not to procure because of innovation on the store for a particular client so you add these two and you will get the big down in this sales on that particular month but you know that the expert that particular you know client is not going to buy that product for that particular month and you have accepted the offer why you should not include it in your time series data modeling say so this way you can modify the four factors and you can integrate all this data some of some this data now this data is data you add or adjust with your time series data here we have made the final decision making you can see this is your basic time series forecast whatever the formula Arima say moving average or any any method you know hold model whatever you follow if it is a seasonal then you follow seasonal forecast also so suppose you found the basic time series forecast and then with human judgment you have adjusted that how come 1483 you found it is nothing but 1106 plus the previous adjustment how much is the 377 so if you add that for the month of January 377 you will get actually 1483 so this is your judgmental forecast say now all this judgmental forecast you have done it now the question is that which is which one is better the mathematical time series forecast that you have learned so far or the new concept that you have added by integrating the human judgment forecast of experts opinion through contextual information and your device forecast that adjustment adjusted forecast so which one is better what you have to do you have to wait for the real data the actually what happened and then only you can see the error in your decision making right in your forecasting process so in reality after one year they realize that this is a case applications you can understand some assumptions are there but you can apply this concept with any practical situation with your real life or the real case study right in India so now reality they found that here the actual sales were these is all predicted model and this is that for adjusted model with human judgment and the other real forecast right look at this September case what happened look at this just one exceptional case I have marked it with a I have highlighted it here what this is that you know here in this case you can see the earlier time series forecast was like this as per the time series forecast say arena model or whatever and now because the one one client said that in the September month of September they will close the store they will not buy because they are going for renovation of the store so you have reduced that almost 300 cells right almost 200 as in 250 cells so you can see the big impact almost 457 amount of in terms of the scaling you can multiply 1000 but here we are assuming that in a basic this data so 457 amount of sales will be down that we have assumed and we have forecasted 1057 but later what happened because in the previous year we have calculated right before the actual business starts that forecasting planning you have done and now practically what happened that client has decided that not to go for renovation at last moment they have changed the policy and they have procured the product so you can see the big spike of your decision making that it failed of your forecasting process so this is the drawback case if you compared it and if it does not happen what could happen that also I have taken into account as a case of example so here you can see that client later decided that not to go for renovation at least for that year and they will continue the store they will open the store and therefore client will come and they have again changed their policy and they said that you pass the order whatever the order I had given earlier so now that has gone up so extra you know 250 amount is come so because of that you could see that there is a big sale here but you know forecasting process revise forecasting process of your human judgment adjustment are not there so anyway overall you can see the modification the reality case and here you can see the error case right the error this minus this and this forecasting error and the judgment minus reality so both error I have mentioned here you can see the big error despite of that you can see the percentage error we have captured in terms of percentage here you can see both time series mathematical forecast and the judgmental forecast and here overall you can see the forecast error you can see this much here in the time series forecast mathematical forecast and judgmental forecast is 97 despite these big cases are involved here big error which we could not predict effectively but if you see the overall percentage error here it is 8% here it is just 6% so still judgmental method is a winner here with this case example but make sure that in future when you study this particular or apply this particular conceptual interrogation of human judgment with the through contextual information feature event into your past data or time series data makes sure you take the weightage part how much impact you want to consider if you consider full of that information like the particular client says that that I will close that particular store because of innovation process for the month of September and what you have decided that what you have taken you have taken the enter information the enter enter weightage that customer actually the client actually give the order for that particular month you have considered in your data if you could have taken 50% of that to avoid the risk part to mitigate the risk part your error could have been further bit less and your accuracy level could have been much higher but you have taken enter delta put 100% of delta so that happened with big error in your accuracy that help that actually made your error forecast accuracy higher but in practice what happens what about the impact or expert are saying how much percentage of that you want to consider that is also very important part so once you integrate once you identify the event and then once you you know capture the the impact of it and how much of that you want to consider the event through these four factors and you can define new factor also not a matter but this as of now let us concentrate on these four factors once you identify the four factors through the event future event and if you can calculate the impact of it as a delta and w the percentage of that how much you want to consider in your forecasting process you may not take the enter portion like if it does not happen so percentage of that you take based on the reliability experts opinion on the confidence of the consulting firm or say expert say and if you can fix these three aspects probably your judgmental forecast will be much better than the time series forecast but if you really if you think that no I will rely only on the statistical data then you discuss that particular session no need to integrate this particular judgmental effect but my point here is that if you can modify your time series data like pre-process as a part of pre-processing outlayer if you can remove because of sudden event it got goes up or you went down if you can remove that and if you can adjust your outlayer why can't you integrate your future event also similar logic because next year after one or two year that will become outlayer and you know that that is going to happen but you are not integrating with the time series data so this particular session helped you to understand how to integrate this one effort you can come up with a better strategy also how to integrate the contextual information through this four factor transient factor transfer factor you can see the graph also here you can see this you can you know mathematical forecast is this one the time series forecast and then say reality is you know judgmental forecast is the rate one and the reality is like this so you can see to some extent I have shown you the error forecast right it is not bad to some extent it might have a you know variation in case it does not occur does not happen but you take percentage of the impact of that particular event error will be less and the measure of accuracy will remain steady or the good as compared to you know basic time series forecasting process so the four factor you have to capture effectively and then the weightage you have to take care into efficient manner based on the experts opinion and how you can integrate it through this or you can adjust your time series process that also we have discussed so this is what on sample illustration of you know integration of expert judgment or human judgment of future events through contextual information and fine-tune the adjust your time series forecasting model and make a revised forecast I believe it is clear to everybody you can practice more at home and you can try of implementing this concept with the real case study with the real situations so let us conclude this session