 Hello friends, we are going to discuss today the various forecasting models which are used to give forecast of the product or the service operations as such. At the end of this session students will be able to interpret various forecasting models. As all of us know what is forecasting, forecasting is the future estimate of the product or services under some specified conditions and for a specified interval of time. There are various methods of forecasting and we will have a close look of the overview of the various forecasting methods. There are qualitative and quantitative models of the forecasting, we are going to discuss those briefly. This is the overview of the forecasting methods. There are qualitative techniques like Delphi techniques and historical methods. There are Navy of time series quantitative analysis like simple average exponential smoothing, moving average, etc. And there are causal quantitative techniques like regression analysis and economic models. Today we are going to discuss the first two that is qualitative and time series quantitative various methods of forecasting. The first qualitative is a Delphi technique. It is nothing but a judgmental forecasting technique where a coordinator poses a question in writing to each expert on a panel. The coordinator brings the written predictions together, edits them and summarizes them. So a team of experts are asked to pose, asked to put their views and their judgment depending upon their experience, the variety of people from various organizations and from various sectors including some customer organizations are also involved in this process. Everybody is posing, everybody is posing a question and he will answer those questions. The coordinator will summarize those and on the basis of summary, coordinator writes a new set of questions and gives them to the expert. Again the coordinator edits, summarizes the answer, repeating the process until the coordinator is satisfied with overall predictions synthesized from the experts. The key of this method is that direct interpersonal relations are avoided, hence personalities don't conflict nor can one have a strong member who generally dominates others. It is generally a case that when your case-to-case meeting is arranged, dominance and then submissiveness are two obvious things. Somebody takes a lead and somebody has to pose a passive role. This method avoids that all the group members are not face-to-face in any of the particular iterations. So this iteration process goes on continuing and tappled down until we reach a particular conclusion. This mainly depends upon the experience and expertise of the people. Coming to the time series analysis, it starts with a very basic simple average method. All of us know simple average. The demand for the periods are equally weighted. So some of the demands for all periods divided by the number of periods. So this is the general formula d1, d2 divided by n. The average reduces the chance of being misled by the gross fluctuations that may occur in a single period. Averages have both some advantages and there are some limitations. Average will try to give the fluctuation to be reduced on a particular line, but at the same time, averages are generally not existing. In any fact as such, those are generally used as a guideline. A better solution for the average remains as a moving average. There are two types of averages, one is called moving averages, one is simple moving average and the other is weighted moving average. A simple moving average method combines the demand data from several of the most recent periods. There are averages being forecast for the next periods. We may use three periods or 20 period moving average. The averages move all the time in the sense that after each period elapses, the demand for the oldest period is discarded and the demand for the recent period is added. In the next calculations, overcoming the shortcomings of the simple moving average. Simple moving average assumes the continuity, but in the simple moving average assumes the continuity. Simple moving average is based on the dynamics of the time. The older data is discarded and the recent demand is added. Therefore, it is slightly better than the simple average method. Further improvement in the simple average method is weighted moving average method. Simple moving average gives equal weight for all periods. It does not discriminate into the latest data and the earlier one, whereas weighted moving average gives you the data which is the most recent and the data which is older one has a different weightage. We have to assign some factor. Therefore, sometimes the forecast wants to use a moving average, but does not have all end period equally weighted. A weighted moving average method allows varying not equal weighing of all older demands. The weighted moving average further, we can have it with the help of a method called as exponential smoothing. Demand for the most recent period is weighted heavily. The weights placed on the successfully older periods decreases exponentially. This is called as exponential smoothing. We are trying to smoothen out the total demand by applying some weight factor. The latest factor, the latest data will have a more weightage as compared to the older one. And therefore, as we go on the data for the older and older side, its weightage decreases and it decreases in an exponential manner. That is why it is called as exponential smoothing. We are trying to smoothen exponentially the data and trying to match with the future demand. The pattern of the weights is exponential in the form and generally the formula for that is f t is equal to alpha into d t minus one plus one minus alpha f into t minus one. Whereas f t is a forecast for a period of t d t minus one the demand for period t minus one alpha is the smoothing constant and it lies between zero and one. Alpha value can be changing from zero to one. It can be 0.1, 0.2, 0.3, 0.6. It depends upon various factors and it also requires some knowledge and experience expertise of the person who is using it. It has to be backed by the statistical method. Now, we have to think that what is the meaning of alpha? What it means when alpha is large? What it means alpha is small? What it means alpha is zero? What it means alpha is one? Yes, I hope now you are ready. You can even think of this particular example by considering your handset for example, you have got a mobile and then now you have to predict the data for the mobile you have to calculate the forecast for the mobiles which you are using and then we have to use exponential smoothing. Suppose you have got data for the past six months what will be the alpha factor you will be applying for all the six months data. So, that should give you a more clear picture of it. It may have suppose some season like Diwali or some other festival season is there how you will apply the weight for that particular season also that you can think of this, right? Well, the answer is as greater is the value of alpha greater is the weight placed on the most recent data. The meaning of it is suppose now we are we want to have a forecasting for a six month period the initial suppose it is the month of now June alpha will have a highest value for the data of May which will have a slightly lower alpha value for the month of April and it will be further reduced in the month of March further we will go on decreasing for the month of February and January and so on. It can be changing from suppose 0.1 to 0.3. Generally most commonly alpha value will range 0.1 to 0.3 in most of the examples. In the extreme cases suppose alpha value is 1. What is the meaning of it? The meaning of it is the latest forecast. Suppose the latest forecast is some X units that exactly matches with the previous period actual demand therefore alpha has come to 1. Suppose the same example of mobiles a particular company estimates that mobile sales for this particular period of the month of July is 10,000 units and the actual demand was also 10,000 units for the last July in that case previous period actual demand is the same. Suppose then we have got a calculation that alpha is equal to 1. The latest forecast we have forecast at 10,000 and previous period actual demand was also 10,000. In this case the meaning of it is that alpha value is definitely coming to the 1. This there is no statistical firm and physical rule that what should be the alpha value depending upon the time period. It is out how that only thing is that exponential smoothing occurs. In the most recent period we will have maximum weightage. In summary today we have gone to variety of forecasting models. It starts with simple average and it starts with moving average. Moving average we place some weighted, called as weighted average and then further so as to be find further accuracy into the data where is in exponential smoothing method where there is a factor called as alpha exponential smoothing its value ranges to 0 to 1. Most commonly it is used as 0.1 to 0.3. Further there will be some more methods which are called as regression methods and all additional methods will be studying into the next series, next lectures. So, thank you.