Demand Forecasting

# Demand Forecasting - Assignment No.3 Demand Forecasting A...

This preview shows pages 1–3. Sign up to view the full content.

Assignment No.3 Demand Forecasting A report Submitted to Prof. Sushil Jhangiani In partial fulfillment of the requirements of the course Marketing-I On 29/11/07 By Aastha Sood Anand T Mayuresh Joshi Prateek Agarwal Sanved Raut Srinath Group 5 Sec D INDIAN INSTITUTE OF MANAGEMENT

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
AHMEDABAD The purpose of the study is to analyse different methods for demand forecasting and choose the method that gives the best estimates for the scooter industry. We also evaluate the performance of the forecasting models and draw marketing implications through relevant macroeconomic events and trends in demand forecast. Firstly, we discuss the different models along with their methodologies and limitations. Secondly, we relate certain macroeconomic events which can possibly explain the trends in the data provided. Finally, we draw implications while comparing the models and understanding if deviations can emerge. Models, methodologies and evaluation: Polynomial Curve Fitting: In this method, we try and fit a polynomial curve to the given data. We try out polynomials of various degrees up to 6 and try to fit them to the data. The goodness of fit is checked using the R 2 figure. The best polynomial for the given data is selected and assumed to represent the given data. This curve is then used to make a prediction for 2004 year of the scooter stock (Exhibit 1). NCAER Method : NCAER used an exponential regression analysis. This is an econometric analysis and takes other economic factors into account. We have calculated the regression between the log of stock of scooters, the log of personal disposable income and the log of the relative price of scooters. In this we assumed the first to be dependent on the last two. The other stock or sale of substitutes was not used since they are not independent variables. The regression line that we got was ln(S t ) = a + bln(I t )-cln(P t ) where S t represents the stock of data in the year t, I t represents the personal disposable income P t represents the relative price index of scooters We found the regression for the entire past data, last 10 years data and last 5 years data. The series using 10 years data gave the highest R 2 figure of 0.94 and the closest match with the last year’s data. The series using all the past data gave a slightly lower R 2 while the one using 5 years data had a R 2 value of only 0.83 which shows large error. Due to the sharp change in trend in recent years we decided to ignore the 48 year data series since it would not be able to accurately forecast based on the recent trends. (Exhibit 5) LPC Method : Linear Predictive Coding (LPC) is a method in which the present value is estimated using a fixed number of weighted past values. The basic equation is x t = - Σ a i x n-i
This is the end of the preview. Sign up to access the rest of the document.

## This note was uploaded on 12/29/2009 for the course OR 203 taught by Professor Vardse during the Spring '06 term at 東京大学.

### Page1 / 9

Demand Forecasting - Assignment No.3 Demand Forecasting A...

This preview shows document pages 1 - 3. Sign up to view the full document.

View Full Document
Ask a homework question - tutors are online