MGMT 361 Lecture 3 Outline of Last Lecture Outline of Current Lecture I. Forecasting-backgroundII. Forecasting Models III. Measuring how good a forecast is RECAP: I. Forecasting Background i. What needs to be forecasted? (EX: Walt Disneyworld)- # of visitors, # of shows/rides to open, # of food/drinks ii. What decisions are made on the basis of these forecasts? - staffing, parking, inventory, pricing, scheduling b. Poor forecasting cost money- in your personal life think: data plan on phonec. Types: i. Qualitative: gut feeling, can not be taught, based off experience, education background, uniqueii. Quantitative: science, can be learned, easy to mimic II. Time Series Forecasts: assume past history is best predictor of the futurei. Forecasting is fluctuating around the mean (for existing products)- most commonii. Fluctuates and increases or decreases as a trend- new iPhone, new products do this b. Basic Times Series: fluctuating around a constant meani. At= actual demand you can observeii. Ft+1= forecasted demand 1. Demand= Mean + Random Fluctuation 2. Forecast= current estimate of the mean c. Method 1: Naïve Method- simple, especially when you do not have a lot of data i. Equation: Ft+1=Atii. Saying that the forecasted demand in the next period is equal to the Actual demand in the period before These notes represent a detailed interpretation of the professor’s lecture. GradeBuddy is best used as a supplement to your own notes, not as a substitute.d. Method 2: The Simple Average Ft+1= (A1+A2…)/t – how you would calculate any average e. Method 3: The (simple) Moving Average: the recent history is more relevanti. Assign equal weight to each monthii. You start using data once you hit the specified “n”- shifts based on n, using most recent months 1. Eqn: Ft+1= (At-n+1+At-1+At)/nf. Method 4: The Weighted Moving Average- weights each of the n most recent demands differently i. Like the weighted Attribute Model ii. Biggest weight to most recent month g. Method 5: Exponential Smoothing Forecast- similar to Method 4 but tell you howto assign weighti. Wt-1= α(1-α)^t ii. Eqn: Ft+1= αAt+ (1-α)Ftiii. *If you don’t have forecasted number for period 1: 1. Use Naïve Forecast for period 2 then go from there2. Use a reasonable number to forecast period 1 iv. We put emphasis on demand so the outcomes are similar III. Mean Demand Level Shifts a. Long-term/permanent shift: give the small n to moving average and larger α for exponential smoothingb. Random event, short-term: give large n to moving average and smaller α for exponential smoothingIV. Measuring Forecast Errors (helps to determine which method to use)a. Et=At+Fti. If MFE is zero, it may not mean that you error is perfect, its actually a poor prediction because there could be a lot of deviation 1. You find the difference, add up the error, and then find to average to get the MFEa. MAD (Mean Absolute Difference)- find the difference between demand and forecasted for each period, to get error, take the absolute value of that error, then add up and average b. MAPE (Mean Absolute Percentage Error)- same as above only you find the percentage or error within the actual demand and find the absolute value of that percentage before finding the mean