 D concepts, welcome to the course on supply chain digitisation. It is jointly being taught by Professor Piña Ka-Biamia, Professor Smita Narayana and Professor Debba-Bratadas his from I.M. Mumbai. In today's lecture I am going to focus on module 3 that is analytics in supply chain management module. So, in this module there are a lot of topics, some of the topics are already being covered. covered. So, what I will do, I will just summarize the topics which we have covered till now and then we will go ahead. So, first we discussed about what is analytics, then we discussed about big data and its characteristics, specifically 6 v is of big data, volume, variety, velocity, variability, value and veracity. Then we also discussed various types of analytics, descriptive, diagnostic, predictive and descriptive and its applications. Then we have seen various examples of analytics from supply chain management domain, then we discussed in detail applications of AI and ML in forecasting and demand analytics. Then we also discussed about supply chain network optimization, how analytics can be used for optimizing the supply chain network. So, in today's session, we are going to focus on intelligent decision tool development in supply chain, like what are the various AI ML based models can be developed which will help decision maker to take better decisions in managing supply chain. So, the idea of today's talk is that we will take case studies and then see what are the techniques can be applied to solve that case study. So, since the topic is intelligent decision tool development, so first we will describe the case in detail and then we will see what are the techniques available to solve this model and then we will discuss one such technique in detail and we will also see how that technique can be developed from the scratch and how that technique will help us in taking better decisions as per supply chain management problem is concerned. So, let us take one such case study. The senior management of a pharmaceutical company wants to expand its business to a new region. So, they have been already operating in a country, but they want to expand to a new region, they want to sell their products to that region. So, they have collected data specifically the location that is latitude and longitude of their 811 prospective customers from that new region and these are the data. So, I have serial number 1 from 12811 and for each customer location we have latitude and longitude. For example, customer location 1 latitude is 27.72290 and longitude is 80.94479. Similarly, for customer location number 811 latitude is 27.51351 and longitude is 80.43675. So, they have collected this data. So, this data will keep them the exact location of each of these customers. Now, the management would like to find out the potential location of the distribution centers because I have to satisfy the customers then demand. So, where should I have my distribution centers? What could be the possible location of distribution centers so that my customer service is improved and my cost is also reduced. So, the supply chain manager is wondering whether she can use any data analytics technique to decide this whole process in a systematic way. That means, where should she locate the distribution centers? How many distribution centers should she open so that customers are served properly their service level is improved. At the same time the whole process is done at a minimum cost. Now, she is thinking whether she can apply any data analytics technique or not. So, these are the latitude and longitude of each 811 customers. So, we have plotted it. So, in x axis we have latitude, in y axis we have longitude. So, all 811 customers are plotted their actual location are plotted. So, the question is how do you find out the potential locations of the distribution center? Where should they locate? If they want to have one DC should they locate it here at the center, if they want to have three distribution center, where should they locate? If they want to have four distribution center, five distribution center. So, where should they locate their distribution centers so that customer service is maximized and cost is minimized. So, let us take an example that they are thinking about opening only one DC. Then would they keep the DC at the center location or somewhere here or somewhere here or somewhere here. So, where should they locate that distribution center? What would be the possible location? And once they decide to locate in a particular location let us see here, then all of these customers have to be served from this DC. So, that is what they have to do. So, that means all 811 customers need to be served from one DC. So, therefore, my customer service time will increase responsiveness will reduce, but the cost of operating DC will be less, because I will be only operating one DC and cost of opening that DC also will be less. Now, let us take an example that they decide to open two DC. So, the first question is where should they locate that DC1, where should they locate DC2? That is the first decision they have to take. So, let us assume that they got to know that DC1 will be located here and DC2 would be located over here. Now, the next question is all of these orange customers will be served by this DC and all of these blue customers will be served by this DC. Now, the question is how do you decide that these orange customers will be served by DC1 and these blue customers will be served by DC2? Let us take an example of these particular customers. So, I have one blue customer, I have one orange customers and both of these customers are located very close to each other. So, now how do you decide that these orange customers will be served by DC1 and this blue customers will be served by DC2? Why can't the vice versa happen? Why can't blue customers be served by DC1? Why can't the orange customer be served by DC2? So, how do you decide, how do you segment your customers? So, that some of the customers are served by DC1, some of the customers are served by DC2. How do you do this mapping? Is there any systematic way? So, that is what the manager is wondering. So, let us think that she wants to define three DCs to be opened. So, the first question is where should she open the DC1? Where should she open the DC3? Where should she open the DC2? This is the first decision. So, let us assume that she decides to open DC1 here, DC3 here, DC2 over here. Now, again the same question, out of this 811 customers how do I segment my customers? Which customers will be served by which DC? How do I do this mapping? So, here if you see the orange customers are being served by DC1, the green customers are being served by DC3 and the vice versa and the blue customers are being served by DC2. Now, again the question arises if I concentrate over here. So, there are few customers. One is orange customer, two green customer and one blue customers. And if I see DC1, DC2, DC3 visually like are very close from these three customer location. But if you see mapping over here, the orange customers are being served by DC1, the green customers are being served by DC3 and blue customers are being served by DC2. But visually if you see it seems that all three are located very close by. So, how do you segment them? How do you decide that this orange customer customers are being served by DC1? This green customers should be served by DC3, this blue customers should be served by DC2. So, that is the question we are trying to answer. How do I segment my customers? So, that from DC to customers distribution happens smoothly. Now, if she decides to have four DC, again the same question where should she have DC1, where should she locate DC2, where should she locate DC3, where should she locate DC4. That is the first question. Once that is answered, second question is how do you map your customer to various distribution center? So, now if I tell you the answer these orange customers are being served by DC1, these green customers are being served by DC3, red customers are being served by DC4 and the blue customers are being served by DC2. So, how do you segment that and say that these should be served by this? So, these two are main decision where should I locate and which customers to be mapped to which distribution centers. Now, if one more thing if you observe, if we keep on increasing the number of DC, then what is happening? Did you observe? So, I will just take you back. If I have only one distribution center, all 811 customers are being served by one DC. If I have two distribution center, some of them like orange ones are being served by DC1 and the blue ones are being served by DC2. If I have three, orange ones are being served by DC1, green ones are being served by DC3, blue ones are being served by DC2. If I have four, orange ones are being served by DC1, green ones are being served by dc3, blue one serving served by dc2 and red one serving served by dc4 and if you notice very carefully you will see that under each dc the number of customers being served are decreasing as you increase our number of distribution center so that means if we increase the number of distribution center my responsiveness will be very fast so I will be able to respond to the customers quickly because this distribution center this particular distribution center is serving only this many customers dc3 are only serving this many customers dc4 are dc4 are serving only this many customers and dc2 are serving only this many customers so per dc number of customers have reduced so therefore I would expect that responsiveness will be much better and customers will be served quickly so if you see the distribution with five distribution center dc1 is serving only this many customers dc2 is serving only this many customers dc3 are serving only this green customers dc4 are serving only this red customers dc5 is serving only this purple customers now if you see six further the number of customers are being served by each dc have been reducing continuously so if I start with one distribution center and keep on increasing the distribution center if I see that results with six distribution centers you can see number of customers are being served from one dc has reduced so therefore I would expect that responsiveness will be good and all the customers will be served from particular dc at quicker time okay so what you observe if we increase the number of dc then we will see that responsiveness will be good because only few customers are being served by that dc so therefore I will be able to reach to the customers location quickly demand will be made quickly so that is the positive thing so what is the negative thing the cost will go up because if I have to open six dc then for each dc I have to incur some fixed cost I have to incur some variable cost so my total fixed cost and variable cost for operating the six dc will go up so although my responsiveness is going up although my responsiveness is going up but cost is also going up okay so therefore I need to do a trade off and find out how many this is should I have so that my responsiveness is maintained and cost is also minimized so how do I solve that problem so that is what we have summarized if you keep on increasing the number of dc then definitely the customers will be served quickly but the cost of opening and operating this dc will go up so how do you decide how many this is should you have and where to locate those distribution centers and which dc will be serving to which customer so these are the few important questions which we are like facing and the supply chain manager is also trying to solve it so is there any systematic way is there any data analytics technique so that we can find out this in a systematic manner so yes fortunately there is a technique called k means clustering okay so we will discuss this in the next class in detail but what we will do suppose you apply this k means clustering technique then how the results will come up and how it will benefit you so let us assume that you are applying k means clustering so in this case k represent number of clusters okay k represent number of clusters okay so in our case if we stick to this example of 6 dc then the value of k equal to 6 okay so I have one cluster which is being served by dc one so I have one this is my cluster one so cluster one that means all the orange customers are cluster one then I have cluster two so all the blue customers are my cluster two so all these blue customers will be served by dc two then I have cluster three so the green customers all the green customers will be served by dc three then I have dc four cluster four so this is my cluster four so all the red customers which are located over here will be clustered together and will be served by one dc that is dc four then we have cluster five so in this cluster I have all my purple all my purple color customers so all these customers will be served by one dc in this case dc five then we have another cluster cluster six so cluster six all the brown customers you can see all the brown customers will be served from one distribution center in this case it is dc six so if you do k means clustering if we apply k means clustering with this data like in which we are having latitude and longitude of each of these customers so if I tell I need six clusters I need to segment this whole customers into six groups so using k means clustering I will get six different cluster so in this case I got six different clusters so automatically your customers will be segmented now on that segmentation is done using clustering technique then the next question is where should I have so once the segmentation is done so if I apply k means clustering with six k equal to six I will get the segmentation customer segmentation now once this customers are segmented then the next question is although I got that these purple color customers should be served by one dc but the question is where should that dc be opened similarly you got to know using clustering technique that all of this green color customers should be served by one dc but where should I have this dc will it be located here will it be located here will it be located here will it be located here where should I locate same thing applies for all other distribution centers so we need to find out where should I have my dc for blue customers where should I have my dc will it be located at this place will it be located here at this corner will it be located here at this corner and so on so this is an important question which needs to be answered so how do I get the answer so can k means clustering technique answer this and tell me where should I have my prospective distribution center yes k means clustering technique will also help you to do that so in this technique there is a concept called centroid there is a concept called centroid so what does this centroid means centroid means the center of the cluster so in this case since the data represents the customer's location so I have latitude and longitude so it will tell me the centroid position so let us take an example of dc1 so dc1 I have all of these customers orange color customers so what this k means clustering technique will do it will give me the centroid value ok centroid value means the center location of that cluster so from that place all of these customers location will be subbed so since this is a centroid location since this dc is located at the centroid of the cluster this orange region so therefore it will help me to improve my responsiveness also so all the customers will be served from this dc and dc will be located at the centroid of the clusters so this k means clustering technique is not only doing the segmentation it is also telling me based on the distance from various customers it is telling me where should I have my centroid and these centroids are location of distribution centers so this dc means distribution centers so these are my prospective location at distribution center so first time doing the segmentation using k means clustering if k equal to 6 I will get 6 segmentation 6 cluster if k equal to 5 I will get 5 clusters so if I show you here k equal to 5 I will get 5 such clusters if I go back further if k equal to 4 I will get 4 clusters if I put k equal to 3 I will get 3 clusters so once I get this cluster then using the concept of centroid I will get the location of dc so these distribution centers are nothing but the centroid of this cluster so dc3 is centroid of green cluster dc1 is centroid of orange cluster dc2 is centroid of low cluster so that is how the possible location of distribution centers are determined so why centroid because centroid is the position from where all the customers will be served based on the distance and distance will be minimum from the centroid so that is the idea so as you have summarized k means clustering is not only helping me to segment my data set in this case segment my customers locations and grouping the customers together it is also telling me where should I locate my distribution center so that the customers are served in a quicker manner so now what we will do in the next lecture we will discuss this k means clustering technique in detail and explain to you like how this technique works and how do I find out the value of k so what would be the optimum value of k will k be 1 will k be 2 3 4 5 6 how many clusters are optimum we will also discuss with you the code like how to write the python code so that you get the same output as you are seeing on the slide so thank you look forward to seeing you in the next lecture.