Lecture 06

Good afternoon everyone, welcome to yet another lecture of the Advanced Business Decision Support System course, which is part of the MOOCS program and I am Dr. Deepak Phule from IIT Kanpur. And along with me in this course is Dr. Amandeep Singh Oberoi and Dr. Pratap Pratap Singh, all three of us will be co-teaching this course together. And this is the precursor of the previous course, basic web based decision support system course that was already been, is not the precursor, successor of the this course which was already floated in NPTEL MOOCS. And if you are not gone through that course, please go through that because lot of the things that we are discussing in this course is part of or is an augmentation of what we already discussed that in that course. And we already seen the previous lectures and overview of the DSS, what are the major components of it. And in the previous course, we have focused more on user interface and database part and this we are actually focusing more towards modeling and how the DSS is applied to various business decision problems, we are talking it more from the business standpoint like a corporate organized decision standpoint. So, without much delay, let us get into today's topic. And today's topic, we will be covering what we call as the basics of modeling and how do we develop models for DSS, ok. So, remember we had earlier talked about DSS as one major component as DBMS, database management system. Then we had a user interface management system UIMS. Then we also had what we call as a KBMS knowledge based management system. And we had MBMS model based management system. So, we are now going to focus in this in the current course, ok. This is our current focus is going to be. We have done this and this as part of the previous course. The major aspects of this was covered as part of this in the previous course, in the previous the basics web based DSS course. So, we are now talking about the modeling and model aspects of it. And we are now in the week 2 of the course, ok. So, this is part of the week 2, what we are discussing today, right. So, our focus is model based management system and for that we need to understand the basics of modeling and stuff like that. So, before we get into the basics of modeling, remember we are talking about a business decision support system, ok. So, most of the time when you talk about a business decision support, the term system is very important to us, ok. And we have to understand what a system is and there are many definitions of system, but the one we are talking about a relevant definition, ok, which what we are going to apply in the business and standpoint. So, we can define a system, ok. It can be defined to be a collection of entities, ok, collection of entities. And what are some of those entities? Entities can be people, ok, machines, materials, etcetera, ok. These are the collection of entities, it can be methods and etcetera. And what do they do? That act, all these entities act and interact, ok. They act among them within themselves and they also interact among themselves and they act and interact together, ok. So, it is in a synergy toward the accomplishment or towards the accomplishment, accomplishment of some logical goal, ok. So, this definition originally, this systemic definition, organizational focus definition and talk about the Schmidt and Taylor, ok, they proposed this in 70s. So, what we say is it is a collection of entities, right. It can be people talk about man, methods, materials, machine, money, mother nature. So, we can think about entities includes 5 M's, ok, man, methods, materials, machines, ok, all that aspect that act and interact together, ok. So, they act and interact together, this action and interaction together. You can think about as synergy, ok, towards the accomplishment of some logical goal, ok. The logical goal can be something like, you know, value, creation, profit, etcetera, ok. So, if you think from a business standpoint, when we talk about a system, this is what we will be talking about, ok. So, this is the theoretical definition of Schmidt and Taylor. So, practically, from a practical standpoint, implied by system, implied by, when you say the word the system, ok, what do you imply by the word systems depends on the objectives of the particular, of the particular decision, ok, or we can think about it as a, or problem, or we can think about as a problem. So, whatever you are talking about the objectives, ok, if you are talking about, for example, maximizing throughput of a production line, ok. So, then you are talking about just that production line. You are not worried about the, you are making assumption that the raw materials and etcetera are already available, you are not worried about the stores and etcetera like that. So, whatever you talk about as a system, ok, in practical sense, this definition of the system can change depending upon what decision or what problem you are trying to solve, ok. So, that is the practical aspect of this. So, we can say we have a factory system, but that does not mean that you are going to make decision through about the factory. You may make a decision about a small component of the factory, but that point, that decision since it is about some small aspect of the factory, the system is only limited to that, ok. So, in another way we can think about it is, so it is important to note this, ok, I am making it as important, ok. Collection of entities, ok, collection of entities that create system, create the system, ok, I am using the word the system in double quotes, system for one study might only, might only be a subset of the overall system for another, for another study. So, what we are saying is that the collection of entities that create the system in one study, a particular study might only be a subset of the overall system for another study, ok. So, depending upon the focus of your study, ok, what you are looking at as the focus of the study of the overall system, you may or may not have the entire system in place, ok, or you may have only a portion of that you are focusing on the study. So, coming to the next part, so then we understand the system state and what is modeling. These are two important aspects, because we are going to talk about modeling in the importance and in the decision support system. So, let us talk about state of a system. This is a very common or very popularly used word, ok. And the simplest definition for it is, it is that collection of, it is the collection of variables necessary to describe a system, necessary to describe a system at a particular instant of time, at a particular time related or relative to the objectives of the study, objectives of the study, ok. With respect to the objectives of the study or the focus of the decision, ok, what do you want to decide? Based on the objectives of the study or the focus of the decision, you require a collection of variables, a particular collection of variables which are necessary, which are required, which are important to describe the system, the state of the system at a particular time, at a particular instance given time, right. Because you can say that the state of system, state of system is dynamic, changes over time. So, if you decide that I made this decision today, 10 years down the road, as the time progresses the system will change, your machines will change, your throughput will change, the capacities will change. So, the decision that is taken today is not going to be valid 10 years down the road, for that matter it would not probably, it would not even be valid after 6 months. So, since the system is dynamic, the business system is dynamic, your decisions will also need to change. So, whatever the variables that is required for you to completely describe the system from the standpoint of the decision or your objectives, that at a given point of time is that is what is called as a system state. So, then what is modeling? So, there are many ways you can look into it. So, let us first think about the academic definition. The enterprise or the exercise or exercise is the enterprise or exercise of devising a, of devising a simplified representation of a complex system, of a complex system aimed at predicting systems performance measures or you can call as matrices. So, you want to measure or you want to quantify the performance of the system or you want to quantify what the system is doing or you want to predict or you want to understand and you want to forecast, figure out what is going to happen. So, the exercise or your enterprise, your endeavor, you are trying to devise a simplified representation. It is not a, so that is why people call it as a model is the easiest way to do it is it is kind of size an abstraction of a system. So, you do not want to study the entire complex system by in itself, all you are trying to do is you are trying to create a simplified representation of a complex system and what is a simplified representation aimed at? You want to study the performance of that system for a particular performance measure. So, in our way such simplified representation, such simplified system representation is called a model. So, when somebody says I am building a model, you are talking about a simplified system representation that is aimed at understanding the performance of a system at point of time. So, also other thing is you should understand that a model is designed usually, a model is designed to capture certain behavioral aspects of the system, to capture certain behavioral aspect or aspects of the modeled system. So, you are modeling a system and so you are trying to capture certain behavioral aspects of the system that you are trying to model, that is the model, the prime aspect of the model is and couple of things are part of this. So, the first one you can think about is these the capturing, what are you capturing? What are being captured? The answer to that question is things that are of interest to the analyst, modeler or decision maker, whatever that is of interest, whatever that is important to the person who is analyzing or modeling or the decision making, that is what is being done. And the second aspect of it is why, why is this done? The answer to that is to gain knowledge and insight, and insight, knowledge and insight into the systems behavior, into the systems behavior. So, the one way to think about it is the rule is you cannot manage, manage is a decision making, manage what you cannot measure. So, by gaining knowledge and insight, what you are actually doing is you are measuring or you are quantifying the behavior of the system, so that you understand the details of the system. So, the other part of it is also modeling calls for two things, one abstraction and number two simplification. Abstraction means only take things that are relevant to the aim or the goal of the study or the decision problem at hand. So, only take that aspects, do not take anything more. Simplification is you may do not want to capture the exact behavior of the system, you may want to. So, like for example, you are just trying to see parts coming into a factory. And if you stand there and notice, let us say for if you take one hour, you saw that that in an hour 6 parts arrived into the system right. So, 6 parts arrive means on an average every 10 seconds, the parts will 10 minutes, one you see one part coming into the system, but in real life it may not be that way. You may see that in the first 15 minutes 3 parts have come in and the rest 45 minutes the rest of the 3 parts have come in, but when you simplify it, you will actually take that the arrival rate of parts into the system is you know 1 in every 10 minutes or 6 per hour or something like that ok. So, that is the simplification aspect of the system. So, abstraction and simplification are the 2 key aspects that modeling actually calls for. So, now before we get into the steps of modeling, we also need to understand the types of the system first ok. So, we can broadly classify the systems into 2 types ok. Number 1 discrete, number 2 continuous ok, there is 2 broadest classifications that are broadly used. So, what is a discrete system? Discrete system. So, the state variables, state variables mean the state of the system, the variables that describe the state of the system. State variables change instantaneously at separated points in time. That specific interest in the points in the time is when the state of the system changes. Let us give it an example as a bank operation. So, the time instant at which the person comes into the bank, that is when the state of the system changes. If I know customer comes into the bank, the state of the system remains as idle typically speaking right. The second one is let us talk about what is the continuous system, I am always worried about the spelling of continuous, cannot spell it properly ever. A continuous system state variables change continuously, they change continuously with respect to time, they keep on changing with respect to time. Let us give an example, the classic example of it is a an airplane flying through there ok. So, every second the state of the aircraft, the position, the velocity, thrust all these aspects keep on changing ok. So, these are the two broad aspects, the discrete system and the continuous system. While the discrete system things change only at specific instances of the time, whereas continuous system continuously with respect to time things change. Now, the we talked about the type of system, so why the major goals of modeling ok. We also need to know why do we do model, there are many goals of modeling, but we are going to talk about from the a business or systemic or system viewpoint. So, we have three major goals, if you look at it from this standpoint. Number 1, evaluating there are many goals, but these are the three major ones evaluating system performance ok or decision process under ordinary and unusual scenarios ok. So, the first one is you want to evaluate the performance of the system, both under ordinary circumstances and unusual circumstances or unusual scenario ok. So, why do we do this? So, one of the aspect is if the routine operation of the system, the routine operation of a real life system, routine operation of a real life system under study, under analysis cannot be disrupted, cannot be disrupted without severe consequences. So, you have a real life system and you are studying the real life system, you are analyzing it, but if you want to do the unusual scenario, you have to stop the disrupt the system performance and that consequences are very severe. So, here is an example of that, attempting production line upgradation while completing orders with tight deadlines, orders with tight deadlines. So, you have many customer orders that you need to complete and the deadlines are tight ok and then you want to upgrade the production line in between. If you do that then the you will have a bunch of pissed off customer at you and that will really cause problem for you. So, this may be the production line upgradation is a routine decision, but then still it is the consequences of that will be extremely large ok. Then other aspect is we said unusual scenarios. The extreme scenario, extreme scenario modeled the one or you want to study ok or studied is to be avoided at all costs. You want to do study the unusual scenario, but you do not really want that unusual scenario to happen ok. So, an example is crash avoidance maneuver, maneuver of manned aircraft. You want to study what the there is going to be a crash and I want to get out of the crash. So, let us me first create a crash and then or create a scenario where crash can happen and then try to get out of it, that let us say you are not unable to get out of it then obviously, the crash will happen and the people in the aircraft will die. So, that you want to avoid at all cost. So, hence you would rather like to create a computer model of it and study it using a simulator or something and then see what will happen if the crash happens ok. So, these are that is the unusual scenario that is the number 1. So, both ordinary and unusual scenarios the system performance in ordinary and unusual scenarios. Second one predicting the performance of the of experimental system design of experimental system design ok. So, the main point here is that when the underlying system when the underlying system does not yet exist model construction and manipulation constructing a model and manipulating it is far cheaper far cheaper obviously, and safer also and safer than building real life systems or prototypes. So, for example, let us say a big vehicle manufacture company wants to really produce create a new factory that can produce 3 lakh vehicles per year and want to know whether such a factory there will be enough demand and all those kind of things. So, one way to do it is the company go ahead invest a large amount of money build the factory make the cars and try to sell it or create a simulation model of a factory where similar throughput is created and demand is simulated and see how many cars are being sold and how many are being held up as inventory or different distributions and etcetera. And then say we actually do not need a capacity of factory of 3 lakhs we only require 1.5 lakhs. So, if you really try to create a factory of 3 lakhs you cannot go back and cut it out and say why only 1.5 lakhs by the time you all invested so much money into this. So, it will be a very expensive proposition. So, hence when you have an experimental design or you have a idea or the actual system does not exist yet. So, if it is not in physical existence then constructing a model and manipulating the model is far cheaper and safer than building the actual system ok. Now comes the third part. Then third one is ranking multiple designs or options and analyzing their tradeoffs. So, you can say it is similar to number 2 it is very similar except the economic motivation the economic motivation is much greater. So, if somebody says I want to see whether annual production of the factory should be 1.2 lakh cars per annum, 1.9 lakh car per annum, 2.5 lakhs car per annum, 3 lakhs car per annum you cannot really build all these factories and decide well the 1.5 is the one and we will demolish the rest of the factories that is a stupid idea and it is very expensive. So, it is better that when you have multiple options or opportunities in front of you to identify which is best suiting for you would need to make build models and manipulate this models to come up with the appropriate decision. So, tradeoff analysis is the third major aspects of modeling. So, these are the these number 1, 2 and 3 are the 3 major prime goals of modeling which are relevant to this course as such. So, now we studied about the types of models ok. So, the first one is the physical model. It is in the simplest sense the definition is a simplified or scaled down physical object scaled down physical object for example, scaled model of an airplane. So, when an aircraft initially is being created or we talk about Boeing or Airbus or something they make a much smaller model of it and they initially they make kind of like a toy kind of a model take it to wind tunnel do the study, then they take make another model put an engine and other things and fly it and try to find out the actual performance of the stuff. And then they finally, go and build the big model. So, it is a simplified or a scaled down model of the physical object that is the physical model. Second is the mathematical or analytical model. It is a set of equations or relations among mathematical variables mathematical variables. So, either it is a set of equations or relations among certain mathematical variables. So, for example, set of equations describing workflow on a factory floor on a shop floor ok or like people say that I have a parcel of land I can either plant wheat or rice and if I plant wheat I will get this much profit, if I plant rice I will get this much profit for planting wheat I require this much of a fertilizer and water whereas, for wheat I require this much more fertilizer and this much less water. So, what is the ideal plantation in my total if my total land area is 2 hectares what should I do kind of a thing. So, that is like simply people will say it as a linear programming problem LP linear problem which you will study in linear programming problem, but the important aspect is that you can create that entire decision problem into the form of a set of linear equations right. So, that will be a mathematical model ok instead of that if you have set of numbers and from there you derive that ok if this much change in x will result in this much change in y can think about that as a analytical model as well right. And then the third one we talk about is a computer model is just a program description of the system of the system when I say program I mean computer program it is more like a logic a logical representation system ok. So, an example of this is a computer model with random elements and underlying timeline is called Monte Carlo simulation model. So, Monte Carlo simulation model is an example of how you can create we can be an equation or mathematical equation, but you really do not know the value of the variables we know that ok this particular variable can take any value between these two in this particular fashion. So, then you use random numbers to recreate the scenario and try to find out how the behavior of the system and how can you model the behavior of the system. So, we can have a physical model is one option a mathematical or analytical model or a computer model these are the three major aspects of model which are important to us in this study ok for this course. The two more other type of models which are sometimes you know is very popularly used and lot of this is to contributed with the help of due to management people and the first one is we call as a descriptive model ok. So, what it does is it estimates a set of performance measures performance measures it estimates a set of performance measures corresponding corresponding to a specific set of input data specific set of input data ok. So, you have a specific set of input data ok and for that input data you want to estimate a set of performance measures corresponding to that specific set of input data ok. So, that is called as a descriptive models example ok. So, simulation models simulation models are clearly descriptive are are usually descriptive and serve as performance analysis models. So, even computational flow CFD models etcetera all are part of these ones ok the descriptive models. Second one is the prescriptive models ok. So, they are geared towards they are geared towards design or optimization optimization. So, when you say optimization what do you mean by optimization? You are seeking the optimal value optimal values or settings ok of a prescribed objective function ok you can call it as a goal subject to set of constraints ok. So, when you start studying linear program and decision trees and etcetera you will understand what we are talking about here, but the idea here is that you are focusing more on optimizing or you are trying to find out the optimal settings or optimal values ok. So, that a prescribed objective function a goal can be achieved within a set of limitations or a set of constraints ok. So, another point you need to note is analytical methods can be effective optimization tools. So, you will see lot of analytical methods used in optimization ok. The simulation based optimization if you want to use simulation models for optimization requires exhaustive search, search for optimum. So, hence and computationally challenging ok computationally challenging. So, analytical methods are better they are far more they are effective as optimization tools. So, if you want to do optimization with the help of simulation then you need to do an exhaustive search and the search can be really you know computationally challenging and both time wise resource wise etcetera. So, thank you for your patient hearing and wish you good luck and I will see you soon with the second week lecture very soon. Thank you very much.