WINTER 2007 MGTSC 352 LEC B1 > DOCUMENTS > LABS > LAB 2 - SIMPLE FORECASTING MODELS Lab 2 - Simple Forecasting Models Course Documents Lab2_SimpleForecastingModels (614.5 Kb) Lab 2 - Forecasting January 19, 2007 Agenda: 1. Forecasting using the simple models 2. Forecasting using Exponential Smoothing (ES) and Double Exponential Smoothing (DES) 3. Performance measures Forecasting using Simple Models In order to complete this lab, you will need the file which contains the daily closing prices of Nortel Networks stock from February 2, 2004, through September 15, 2004. The purpose of this lab is to review the material covered in class with an emphasis on the forecasting tools that Excel offers. Scenario Way back in the year 2000, your Dad received a hot stock tip from his buddy Larry. He trusted Larry so he bought a pile of Nortel stock for $115 a share. Soon the stock started to plummet but your dad held on to his shares hoping that he could recover some of his investment (Larry's body was found floating in the North Saskatchewan a few months ago.) He rounded up what was left of his life savings and has paid for you to go to University for the last few years but now he needs some more money to help you in the new semester. He heard you were in a fancy forecasting class and he would like you to help him predict the stock price. Since your future at University depends on this money, you decide you better pay attention to your forecast-loving instructors, Morgan and Jen..... In order to complete this lab you will need to use the data on the Nortel worksheet, which contains the daily closing prices of Nortel Networks stock from February 2, 2004, through September 15, 2004. Given such data, how would you forecast the closing price of the stock on September 16, 2004? Forecasting 1) Last point: Ft+1 = Dt; 2) Historical average: Ft+1 = average(D1 , D2 , …, Dt) 3) Simple moving average (SMA): Ft+1 = average(Dt-m+1, Dt-m+2, ..., Dt), where m="window", suppose m=4; 4) Weighted moving average (WMA): Ft+1 = w1 Dt-m+1 + w2 Dt-m+2 + …+ wm Dt, where the weights satisfy w1 + w2 + …+ wm = 1, suppose m = 4, w1 = 0.1, w2 = 0.2, w3 = 0.3, w4 = 0.4; 5) Exponential smoothing (ES): F2 = D1, Ft+1 = LS Dt + (1-LS) Ft, where 0 LS 1, suppose LS=0.25; Initialization: Learning: 6) Double Exponential smoothing (DES): Ft+1 = Lt + Tt. Lt = LS x Dt + (1LS) x (Lt-1 + Tt-1) and Tt = TS x (Lt - Lt-1) + (1 - TS) x Tt-1, where 0 LS and TS 1. Suppose LS=0.25 and TS=0.25: Initialization: LEVEL... TREND...... Learning: Prediction: 7) Which of all of the above forecasting methods would you recommend to your Dad? You may want to compute some performance measures before answering this question. Error Calculations: Error: the difference between the actual data and the forecasted data Absolute error: the absolute value of the error Percentage error: the absolute error divided by the actual data Square error: the error to the power of 2 Performance measures Note: n = number of data points one can calculate errors for BIAS (average error): Compute the average of the errors. MAD (mean absolute deviation): Calculate the average of the absolute errors. MAPE (mean abs. percent error): Calculate the average of the % errors. MSE (mean square error): Calculate the sum of the squared errors and divide this by the total number of errors less one. SE (standard error): Calculate the square root of the MSE. Take a look at the other methods for calculating the performance measures. These can be found on the completed tabs. Feel free to use any method you're comfortable with. That's the nice thing about MGTSC... there are often many ways to come up with an answer!