 Hello everyone, welcome back to the session of Data Different Decision Making and Essentials of Predictive Analytics. In the previous session, we have discussed different type of business analytics or say data analytics and then the steps of data different decision making. We discussed the descriptive analytics, the predictive analytics and the prescriptive analytics also. We also discussed that what are the essential ingredients or the techniques that you need to know to understand or to become an expert on the data different decision making domain and also we have discussed to some extent different requirements for descriptive analytics, predictive analytics and prescriptive analytics. We have also discussed the data different decision making process like business understanding, data collection and pre-processing, model building, model evaluation, model development, model maintenance, etc. We have discussed detail of business understanding with different steps and the requirements and also to some extent essential of predictive analytics also to some extent we have discussed. Now we will be entering more of or to some extent in depth of data, pre-processing and the collection of data types so that you can enter into the model building on what are the techniques available for model building process of especially the predictive analytics models. Then model evaluation requirement and model maintenance part we cover. Now let us come to the data collection and the pre-processing process. Generally you know in when it comes to the data collection you need to think about the data source, the reliability, the consistency of the data, the accuracy of the data and like you know to some extent biasness or outlayer you have to think all these things. So reliability and the source of data and the data consistency is extremely important before you make any informed decision by analyzing the past data. So therefore accuracy or ensuring the accuracy of the data, completeness of the data if the data are not available I can give you many such examples where we fail to develop some new innovative models because of the non-availability of the data or reliability of the data. I will give examples in some session you will get to know like you know in India there are many government sources or you know many you know practical platforms where data are not completely available or completeness of the data are not been there. Therefore if you want to do some any project or some you know some research activities what happens you know you might not rely on that particular data and hence you will fail to not make better informed decision based on the data available data with you. So therefore you need complete data, reliable data and consistent data so that you can make a better accurate forecast or you can say you know predictive or prescriptive decision making or business analytics modeling. Then the quality of the data are also very important you might have the available data or reliable data but the data which actually make help you take a better decision or you know to some extent you know improve the accuracy of your predictions may not be that quality of data is also available. So you have to think about the quality of data there might be very data which are very vague or to some sort of fuzzy type data which may not be suitable for your decision making. So therefore you have to think about the quality of the data also. Also when it comes to the source of data make sure whether data are continuous data or discrete data whether it is a qualitative data or quantitative data we have a session inter session on this particular slide itself like you know type of data and the you know different type of modeling process of data type also. Inter session are been you know been allotted for this particular information that understanding the data type and you know what type of you know modeling approach you can you know apply for different data pattern that also we will discuss but here I am giving a summary as a part of introductory session of essentials of predictive analytics. You can think about whether the data are qualitative data or not quantitative data or not. If the data is a qualitative what kind of you know subjective approach you need to follow. If the data are qualitative what kind of group decision making approach you need to follow if the data are quantitative in that case what kind of regression analysis or what kind of causal models or say time series models you need to use that also you need to understand and accordingly data patterns understanding reliability consistency data you need to understand. And corresponding model you have to bring of typical quantitative analytics of quantitative modeling approach. Then whether data are structured or not unstructured or not that is also very important in terms of text mining or the machine learning process you need to really require whether data are structured or not or unstructured data whether if you have a data of this type how to handle this type of data also you need to understand whether data is a big type or big data type or not. If the data are a big type how to handle all these things you know whether 3 v are involved over there or not data are not in large in volume still you can talk data are a big data also if the variety and the velocity of the data are quite high. And therefore you need to understand the data pattern effectively and the corresponding collection like cross sectional data next slide I will discuss about that cross sectional data time series data panel all these things are very important about the data type. Once you know that then you know you need to think about you know data collection process also like whether the primary it is a data you need to collect through primary sources or you need to collect the data from the secondary sources. Then the major important part you know that ethical consideration you cannot data collect the data which are not been allowed to you know take without permission from the organization. So, you need the ethical permission you need the safety that safety you have to provide that you know declaration to the organization that you know you will maintain you will not use for commercial purpose or whatever the region it is been given to you you have to you have to maintain that protocol guidelines safety of the data and the privacy of the data you have to maintain the legal compilances are also very important to during data collection process that especially the ethical consideration. You have to you know declare the ethical parts about the data collection and the you know maintaining the privacy of the data. So, these are the you know couple of important aspects that you know before you enter into the data collection process for any research or any study you have to maintain this particular couple of aspects of data type and the data collection process. Then I was talking about primary data and the secondary data. You know more detail we will discuss in the data type session or type of business forecasting session. So, but here as a summary let me share that two type of data collection process one is the primary data collection or the primary type of data and the secondary type of data. There are two major aspects of data type and the data collection process one is the primary that is called you know through survey through focused group interview process or you know data collection through Google's seed or so you know through social media network you can collect the data through alumni network also you can collect you can directly visit to the consumers you know stores and you can take the opinion you can directly visit to if it is a manufacturing related projects to you have to collect you have to visit to the you know manufacturing plant and you have to take the opinion of the managers the labor etc. Then only you will be able to take a decision over there. So, all this may fall under the primary data collection or through survey process you know you can contact the interview also for qualitative analysis there are very good techniques involved through you know like you know institutional theory, stakeholder theory through that you can also you know do some primary data analysis also collect the data through interview process you can do the sentiment analysis also of the interview and then extract the data through any software there are many you know techniques are there and then you can take a informed decision through data analytics techniques. So, this type of process you know social media monitoring also through that also you can do there are many marketing agency available in the market through them also you can collect the data whether from the top manager managers opinion you can also take you can take the you know direct consumers opinion also all these things fall under the primary data collection through what about the way you follow like for example here I have given an example that if you want to identify the significant factors that you know impact the last mail delivery of say you know any supply chain or the e-commerce domain in that case what are the important factors that you want to study what you have to do you have to create questionaries by literature survey you can identify the factors and then you have to validate that for the expert opinion right who are working on the last mile domain of supply chain or say you know e-commerce domain so that take their opinion and then you you know clear create your Google seed or you know direct interview and that survey process fall under primary data collections but when it come to the secondary data collection for the same problem I talked about you know say last mile delivery etc in that case suppose if you think about the e-commerce domain same domain I have kept as examples here you can see it's a secondary data so data secondary data says that data which are been stored or been you know be developed or the decoded by someone else when you collect that data from someone else through secondary person so we call it as a secondary data or like you know public or existing information collected by others you are taking that data and you are analyzing that called secondary data you know you go to Demat store say every day huge amount of data been stored right in Amazon every day you are giving order every people are making order through Amazon say or say you know Flipkart etc or then then all the data are been stored over there so you are not collecting that data they are collecting that data and you are collect taking that data from their sources through their SQL database or you know DBMS database and then you are making your decision making through data analytics steps techniques predictive analyzing modeling so that data source of data or that the collection of that process fall under the secondary data analysis or secondary data collection process it can be in a company websites it can be government of statistics it can be in a industry associations you know publicly available market research report or you know different database sources also the example that I was talking about last last mile delivery in that case suppose you want to see that how many products are returned by the consumer suppose consumer give the order and they keep the product in their cart or say you know cancel the product even after delivery also in that case how many product are getting cancelled that actually hinder the business of the e-commerce firms so in that case in the last mile if they want to focus that how I can improve my cancellation for the consumers or how can improve the customer satisfaction or hit rate in a better manner so in that case you do not require to some extent the primary data analysis you need to see the data pattern the consumer behavior of your data how they are which product they are canceling why they are canceling all this you know like you know in which day they are canceling the reasons of cancellation suppose or return of the product you might get to know once you store the data of secondary analysis entire product characteristics consumer behavior etc you can study based on the secondary data analysis that falls under you know second category called the secondary you know data so there are two major category one is the primary data analysis or primary data collection process the secondary data analysis or the secondary data collection process then all this I am talking about the data source data type and the data collection process right now if you think about the data type here I have listed you all are aware about all these things but I have you know categorically mentioned all this type look at the examples and the type of mean analysis like regression time series analysis which we will discuss in detail we will get to know then the categorical like you know data say you know red female you know male etc blue etc different type of product in orange apple this type of categorical if you define that then what kind of method you can follow national learning techniques you can follow whether it is a logistic regression or the classification of the decision tree analysis you can follow that look at the ordinal number you know small medium large category and then rank order analysis you can use look at the binary data yes no kind of thing whether you know the product will be like insurance will be provided or not loan will be given or not so in that case you can take a decision by yes no kind of thing to false kind of decision logistic regression of hypothesis testing can also follow contextual data like you know text mining you can do like you know how hello how are you different like people are you know suppose something some movie are getting you know popularity in the market so people are tweeting on that or people are putting comment on the social media so if you take that social media data and then if you do mining over there and you get sentiment over that the customer sentiment etc on the data then you can follow some text modeling approach also no then you can think about say spatial data so in that case you know you can use GPS coordination you know also you know like in a map type of examples in that case you can use GIS technology think about you know time series analysis time see if the data you have time series temperature also in stock price etc you can use forecasting on the trend analysis image also here I just talked about you know you can use about so much photography extra images etc in that case if you have a that type of data you can use image recognition or you know computer vision all these things processes network analysis if the data is of that type like you know draft theory techniques or you know network analysis you can use also audio type data also you know in that case in a music is speech recording also like you know if you do a primary interview with the data with the experts so in that case you know your data might be an audio type right so in that case you can use some speech recognition software audio signal processing like I was talking about in the software through that you know you can extract the speech recognition through quality forecasting methods or quality to you know analytics methods so these are the different type of you know data and couple of examples I have mentioned and also some techniques I have mentioned for you whenever you required all these things type of data and which techniques to use you can refer this particular slide and you can you know think about understanding the methodology of the techniques and you can implement that particular analytical models or the data analytics techniques then next part which are very relevant to our courses of business forecasting that is called whether the data are of stationary or non-stationary data I repeat all these I will be discussing detail in data type session but here let me give a summary of definition these two station of what is the difference between stationary and data analysis non-stationary data they are very crucial if the data are of steady over a period of time or you know variation are not been there involved over period of time like you know mean variance and the standard emission of the data I will not change much that I will show you in the next slide in the graph so that type of data are called as a stationary data and it is easy to make a forecast for the future but if the data of non-stationary that means it is too much of jigsaw variation or randomness are involved over there so in that case you know your mean of the data for a different block or different time period will shift so if the mean on the variance etc changes over a period of time in that case that data are not stationary we call them as a non- stationary data it is very difficult if the data might have up or train down train and then again down train up train etc seasonality also many factors can involved over the data random behavior like you know behavior can also come to the data in that case it is not easy to make the forecast for future because what has happened that may not be followed in the future so in that case how to make the forecast so that type of data analysis is very crucial for you know business forecasting for the time series analysis we call them as a non-stationary data here one is a small example that I have mentioned look at this data are called as a stationary data because any block of the data over a period of time if you take you know what happens you know the mean of the data remains same almost it remains same it not changing much mean variation so the data remain inside a you know you know range or you can say inside a you know band line so it is very easy to make a forecast for the future also but look at this data the mean may be here but once you shift the block from this location this time period to that time period you can see the mean has shifted so to some extent there is a big gap in data pattern in later in mind go up go down you know so in that case it is very difficult to make the forecast for the future so how to confirm non-stationary data into stationary and make a informed decision will study that through different modules of you know business forecasting then you know another concept also you should know that is called the whether the data of time series data or the cross-sectional data if data are been collected you know at a same time period for different location for different variables that type of data are called the cross-sectional data but you collected the data for a same time period look at that you know cross-sectional data consist of observation of many subjects at the same point of time but when you talk about the time series data use you stick to a particular parameter particular variables and then you collect the data over a period of time over a period of time you collect the data for that particular variables and the way then you analyze that data for you know future analysis or the future predictions so that data are called the time series data which are been recorded or the collected maybe primary or secondary process whatever the collection process you follow but will be collected over a period of time that is called the time series data but cross-sectional data are at the same time period look at on the other hand cross-sectional data focus on the several variables at the same point of time so same like today suppose you want to collect the data say temperature or say humidity or wind speed of Delhi, Bombay and say you know Calcutta, Mangalore, Chennai at same time period you check what is the temperature there what is the air quality etc you check and then you make a decision over there that type of you know you can do many analysis also like social analysis also social science project you can take and you can do analysis over there for different city demographic analysis also you can do that type of data may follow under cross-sectional data so you have to be very careful about what type of data you are handling and what kind of you know panel of data you are managing and you are making forecast for the future or analyzing the data through data analytics techniques and tools now let us come to the data pre-processing so here you know remember it once the collection process are done or you know data understanding are been over then you may make sure that data should not have in out layer data there should not be missing data if there is a missing data how to come up with the appropriate replacement of data or how to remove that data that you have to take care also data transpiration you have to do scaling also like you know you can take Laplace methods or you know you can take the logarithm to scale down the data you can reduce the dimension of the data so all this are part of data pre-processing so missing data out layer you know removal of duplicates all this you know string variables analysis all these things comes under the process of data pre-processing which is a part of descriptive analytics or you know exploratory data analysis very crucial this part is that data pre-processing if you are strong enough of handling a school data analysis through different techniques of you know data pre-processing probably you will get a clean data and also you can make a better forecast like you know you can think about say skewness let us think about say you know measure of like it is all about exploratory data analysis right suppose one examples are given here skewness look at here you can think about data this data has a symmetric distribution because the mean median mode are falling in the same point but look at here this data has a little bit of skewness because if you change the data time collection or the you know data variation changes over there so you can see that the mean has been shifted here so therefore it's to some extent positive skewness look at how this data has a negative skewness so means has come down to the negative side or you know you can think about you know this type of data analysis very carefully you have to analyze the data there are software there are techniques the last module of this business analytics session I will show you on software through that you will get to know that different type of data analysis inter-exploit data analysis will get to know it in half an hour so all these you know calculation process or the data pre-processing how we can get to know one at a you know step I'll show you through you know excel at advanced excel software called address software and then you can think about say you know look at the pattern of the data whether you need to read by read multivariate data behavior of pattern here I have shown you on data behavior of pat or say pattern of the data you can think about that red color data which are now the suppose a actual collection of the data or say in a sales of the data but it is following some you know uptrend of the data as well as you also it is following some signality pattern look at the pattern of the data it is following some peak at particular time period of a year particular time period of a year look at here so here you can see this I have already mentioned here you can see all these peaks are nothing but some signal variation it has been repeated every time period but that will follow similarity in the fourth term here so how to make forecast which model to select so all these things are part of you know explorative analysis and the appropriate selection of the model so that look at here randomness randomness is also involved so suppose COVID during COVID there might be cases that where there are no sales stores are closed so in that case this randomness you cannot bring in the next stage here also right your system has come back to the normal situation now so in that case how to analyze this data how to remove this you know outlier or bring the smoothness of the data that cleaning process the explorative data understanding the data pattern and all these things you know are very crucial for prediction modeling or say descriptive modeling so these are the initial part of you know descriptive analytics or the data preprocessing and to some extent data collection and the data modeling part as a part of descriptive analytics or model building process stage 2 of model building process under the data decision making now next step is the look at the first step of data decision making that is called the model building right under model building the major concepts of you know machine learning or the predictive analytics will come even prescriptive analytics will come all these things are part of data decision making the major part they are the model building nothing but are the predictive and the prescriptive analytics under predictive analytics you can think about different type of supervised learning but you know also you know some there are many techniques of model building like you know dimension reduction or you know factor analysis all these things are part of you know network analysis are part of unsupervised learning also look at here I have written couple of steps of you know unsupervised learning there is a session called introduction to machine learning under this business forecasting course I will discuss detail of this particular slide especially this part of you know unsupervised learning and the supervised learning techniques and the application process of that but let us see the summary of this like you know clustering process segmentation k-means clustering you know hierarchical process of data segregation and then in a linear dimension reduction support vector machine you know principal component analysis they are very important for example support you have a many dimension of the data right many variables many parameters other but you cannot handle all them but you might have to take the most appropriate parameters of the data and you can show that you can understand the data pattern and you can make the better forecast for the future in that case you might reduce the dimension of the data through say principal component analysis I can give one example says for example say suppose you have a say logistics cost right so you want to minimize the logistics cost you want to optimize the cost you want to optimize the time you want to optimize the route so all three are optimization right route optimization time optimization cost optimization so there are many parameters but you can give a small name like you know latent variable name say which may be a optimization only so that optimization of the hit rate optimization may you know meet all these parameters together you can reduce the dimension and you can make carry forward the discussion of the data and analysis of the data and you can take a better decision so there are many way of you know reducing the data pattern other data like dimension reduction factor analysis also many factors may be relevant to your model right but you cannot take all factor you have to take the most important factors to take forward with the decision making or the analysis further so in that case which factors are more important we can validate that through the experts opinion also so we call them as a you know to some extent factor analysis process also then association rule and recommendation system very crucial in marketing domain actually or you know in IT industry like you know or AI process of association rule like market basket analysis all these things may be you know put under the umbrella of unsupervised learning network analysis text mining which falls under the unsupervised learning of machine learning aspects of predictiveness then when it comes to the supervised learning or predictive modeling which will focus majorly in our courses of business forecasting here you can think about you know classification of models like logistic regression decision tree analysis random forest you know all these are very important look at the regression analysis simple linear regression multiple linear regression will discuss detail of all these things in the session of you know regression analysis module will discuss moving average models exponential smoothing models look at the I have mentioned about exponential smoothing here but effectively under that also there are many models like hold model winter hold model decomposition method all these things will come under exponential smoothing concepts itself will discuss them will discuss also Arima model SCF autocorrelation function partial autocorrelation function auto regressive integrated moving which is most powerful techniques of predictive analytics or machine learning process you study that also to some extent will cover the basic information of the you know different type of machine learning technique like bagging boosting and also you know to some extent support vector machine also or since there are these are couple of models like neural network I will not cover the neural network part in the detail of support vector machine but the introductory aspects of all of them you know unsupervised and supervised learning process I will discuss under the session of machine learning technique as much as possible will try to cover this part then there is another process of machine learning or to some extent you know AI process or you know data science process called the prescriptive analytics. I have discussed in the previous you know session of you know this data decision making session that what are the techniques involved under prescriptive analytics like optimization technique mixed integer programming problem non-linear programming Monte Carlo simulation decision tree analysis Bayesian analysis utility theory they are very important techniques steps of you know consulting or you know statistical decision making we call them as a prescriptive analytics module and then you know game theory multi-creator decision making MCDM technique are very popular in industry in decision making process ranking process and different type of statistical inference all these falls under prescriptive analytics. So, there are three major aspects one is unsupervised learning or machine learning or predictive modeling and then the you know supervised learning of predictive analytics which will focus more on that and then there is a third context that is called the prescriptive analytics which falls under you know the business analytic aspects. So, these are the you know general aspects of you know or the techniques that you should know to become expert in data scientist or the two experts of business analytics domain which you can put all under the model building process. Any software you bring they for throw them whether it is you know advanced excel or the you know python mad layer other advanced software where you are people are people are you know are say people are using this type of techniques through different algorithm and they are addressing practical or cases or practical problems. Then the next step are the model evaluation. Model evaluation is nothing but once you develop the model with the data and you are taking a informed decision through strategy in that case you have to bring the accuracy of the model or the reliability of the model. You can take any model and you can make a forecast right suppose you have a data like this you have made a forecast like this whether this forecast is best or not that you have to you know get to know through your measure of accuracy. So, there are different type of measure of accuracy techniques are there there is a session where we will discuss detail about you know different type of measure of accuracy with numerical examples and the illustration will study all these techniques of measure of accuracy like mean absolute deviation, mean absolute percentage error, mean square error and the root of mean square error like RMSE in financial sector is very popular. So, we will discuss all these things on that particular module measure of accuracy module, but in general what happens now through this model you know evaluation process you need to measure the reliability of the model the accuracy of the model whether this forecast is best or that forecast is best. So, we have to bring the calculate the measure of accuracy and according to you have to make a forecast or comparative analysis that whether your model is best or not like this with the data trial data and the testing data you have to bring the accuracy in the reliability of the model we call it as a measure of accuracy there are many like in regression analysis we use you know R square right in time series analysis we use mean square error even even you know in some logistic regression model or some decision making we use you know RMSE or say you know you can say you know say you know accuracy level between the actual prediction by the total prediction. So, you know accurate prediction by the total prediction this way you can calculate the accuracy level in insurance sector you know or banking sector people use this type of evaluation process whether your prediction model or the categorical model or some different type of logistic regression model that you are using whether your accuracy is above 90 percent or not or how much is the demand threshold requirement that company has provided to you have to maintain that. So, there are different type of you know measure of accuracy other model evaluation process are there you know confidence interval correlation coefficient you can use all this you know like utility of money you can also put as a measure of accuracy and you know different type of optimization threshold value also for them you know expected monetary value also you can use and you can get to know that whether your model that you have developed is really good or bad or how much reliability has compared to your computer or peers. So, these are the process of model evaluation in each and every model you need to evaluate the accuracy of the model through a couple of these techniques that we learn also. Here is another example you can think of the clustering process you can use the clustering of unsupervised learning and look at the data here you may not be able to take any decision right. Once you use the clustering process you may classify the data into different category and you will be able to take a decision whether you know whether the old age people are inferred or not who is buying your product middle age people or the you know to some extent you know younger age all these things you may get to know different type of classification you can do through clustering process came in clustering types and then you can take a informed decision over there which is also part of model evaluation as well as the model building process. So, just couple of points I have mentioned here more detail we will discuss when we will open the sessions then once you evaluate the model develop the model evaluate the model then make sure that you have selected the appropriate model there will be many models which you will study right and you will get to know about the application of them but at the same time you have to understand whether you have understood the context of the model or the context of the data effectively or not and the reliability of the available data there or not because you are taking a summary now you are trying to implement the model right you are the deployment process started now. So, in that case you have to think about the degree of accuracy the desired like the reliability that just now we have discussed about that right the measure of accuracy also forced cost accuracy also time period to be forecast that also very important because now you are going to implement the model to your organization with the data that analysis that you have done another value of forecast to the company how much requirement whether you are meeting the goal of the company or not the objective the research question that you have said that you are meeting it in or not that you have to think also also the stage of product life cycle this is also very crucial for example for startups in the early stage the decision making may be different the forecasting accuracy may be required in a different manner or the you know decision making analytics may be different, but when it will be the existing stage also growth stage the decision making may change the forecast accuracy may change risk measurement may change. So, in that case your analysis also will change your forecast accuracy will be also varying. So, in that case you have to think about at what stage of the product life cycle also time available for making the analysis also it is very crucial if you have a huge amount of data or in a longer period of data you can make a better forecast or you know if you if you have some you know short term prediction or long term prediction that classification you can do and then or based on the availability of data you can also take a decision. So, the time assessment the time availability of data and the requirement for forecast whether the short term prediction long term predictions they are also very crucial. So, therefore if you know these particular you know 7, 8 steps probably you will be able to select the appropriate model that you will learn and you can make a better informed decision through data analytics steps and the processes. So, once all these are done next is the deployment which I have discussed that you know intricate the model the techniques that you have developed to the company and develop a software decision support system or server kind of system where you know you can put the data and you can make a you know database system or you can you know which will reduce your cost increase the revenue bring the operational efficiency you know and improve the customer satisfaction. This is a major step you know by by deploying your model effectively you are solving the case and you are making a good recommendation to the organization that you know follow this prediction model, follow my recommendation and you can automate your decision making you can optimize your business process you can enhance the operational efficiency and increase the market share or the customer experience. So, this is what called the model deployment. So, once this model deployment part are done look at that you deployment of model can be utilized repeatedly on new data remember this part the next step will come here like maintenance or this is called commissioning once this is done you have you cannot conclude the inter decision making process next is that deployment of the model can be utilized repeatedly for no new data making analysis process reproducible and adaptable to changing business requirements with the new technology complex chart, GPT etc. you have to modify your AI process right you have to modify your data analytics techniques. So, once you get more data you might see the data of customer behavior are changing now computer has come back has come to the system you have a many product into the market. So, how you can make better decision based on the availability of data with the new data or customer changes you know by capturing the customer changes the experiences. So, in that case whatever you have developed that may make an absolute unless you modify your data or modify your model and again you redeploy your model again. So, that calls actually you know model maintenance it is very crucial actually model can be affected by data drift and the concept drift remember it if you don't manage the data with and understand the business dynamics over a period of time the changes of customer test and the you know different situation competition etc. if you don't capture that the example that I was giving in the first session of this particular data decision making session think about the example of Nokia think about the examples of you know say or could also Yahoo etc. think about you know the current demand of say Facebook etc. think about you know even Facebook also you know changing with their decision making with the more data and they have entered the metaverse domain think about you know the examples of you know say Geo entering or the Geo into the you know telecom sector and what happens to the both of them they may come back again with their strategy etc. so therefore model maintenance and the deployment is very crucial if even both of them Geo both of them and say idea if they come up with a better analytics process better customer changing policy strategy with a better financing process probably they can also increase their market share so therefore model maintenance is very crucial apart from model deployment so updating your data revising your model and making a better decision or revised decision redeployment of the model is called as a model maintenance or to some extent monitoring and control of the modeling process and the decision making process remember whatever we have discussed the steps first we have discussed about you know data understanding or you know context of the business problem then we have discussed about you know data pre processing model building model deployment and model maintenance if you know this part through three aspects of data given decision making or the business analytics descriptive predictive and prescriptive or to some extent you know aspects of business analytics probably you will be able to you know overcome the business challenges and gain better competitive advantage. This is the summary of so far that two sessions that I have discussed in the previous session and today's session this is a summary actually overcome by understanding data decision making or understanding the basic aspects ingredients of you know essentials of business analytics or predictive analytics probably you will be able to overcome the business challenges and gain the competitive advantage. This is what the advantage of data decision making. If you think about application of business forecasting or the business predictive modeling techniques that you are going to long through this particular courses you know it's it's not limited to this particular you know list it can be applied in detail sector data analytics and the business forecasting and the business prediction modeling are need of the hour actually. So therefore in detail sector also you can think about application of that in manufacturing sector you can use right whether you know focus the production level raw material requirement energy consumption or say you know production planning system everywhere you know you can use this data analytics and so predictive modeling over there also. In detail sector I don't have to give everywhere you know you you require data analytics. In financial sector in banking you know entities are actually implementing the data analytics over there. Even SBI also taking initiative the data analytics for SBI kind of you know like in a branch they have already initiated their branch. Even you go to the insurance sector you go to the NBFC sector everybody is adapting the new technology the data decision making to make their better forecast informed decision making in the financial sector. Healthcare sector the perhaps this is the only sector where you know people are not focusing much on data analytics. Perhaps the most important sector that one should focus or the government or the organization should focus that they should integrate the data analytics in healthcare sector. Perhaps this is the only sector where there is a huge opportunity to explore the integration of data analytics or the business prediction modeling in that particular sector so that they can take a better decision for the consumers for the customers for the patient and for the organization also. Transportation sector also even this they are not only the data analytics people are using you know blockchain also in transport or IoT system also. So therefore transportation system also have a better impact like you know forecast traffic congestion. Imagine you know like you know we will all talk about you know Google map right and and also we are like GPS all these things we see in our mobile when we try drive our car and all these things. But you know most of the Indian companies even also are following are not following the Google product they are actually following some you know map my India product to you know assess the route through their transportation process. You know we are also perhaps following that. So this is one of the Indian company map my India and that is they are using data analytics so effectively that most of the companies are not taking that like advantage of Google's map they are following this particular map my India's product software and they are implementing in their you know tracking system of transportation system. So imagine this is what advantage if you if somebody follow even small entity follow some data analytics and the corresponding you know informed decision-making probably they can also be the winner. So here I have given one example in energy sector also you can think about you know it's a big demand that inter-enabled energy sector are evolving as a future of country our country or I mean throughout the world so therefore you know integration of analytics can be a big boost to that particular sector. Logistic sector I have already given examples the technical telecommunication sector, media, government everywhere. I as I mentioned these are not only the couple of sector that where you can integrate in which sector you belong in which sector you are going to be data scientist or in which sector you are going to be a consultant not a matter but if you have a expertise in the data analytics and the application domain in this particular couple of sectors probably there will be age in decision-making through this you know courses of business forecasting or predictive modeling. We will try to cover as much as possible all this you know application domain also through when you open different modules of business forecasting or predictive analytics we will also try to give examples from the all these sectors so that you can match or you can map or you can get to know better mapping part between the techniques that you learn and also the application part. So, through the case examples through illustrative examples I will try to cover this particular application aspects also so that that the mapping between the industry requirement and the academic understanding of the technical aspects should the gap should reduce and the you know better insights may be you know passed to you through this course of business forecasting and very essential of predictive analytics. So, with that let us conclude today's session of essential of predictive analytics and data-definition-making steps. Thank you.