Forecasting Professor Ahmadi Slide 1 Learning Objectives n

Forecasting Professor Ahmadi Slide 1

Learning Objectives n n n n Understand when to use various types of forecasting models and the time horizon. Understand qualitative and quantitative tools of forecasting. Compute moving and weighted moving averages, and exponential smoothing time-series models. Analyze trends and seasonality in time-series data. Compute variety of forecasting error measures. Use variables in linear regression model. Use Excel to analyze variety of forecasting models Slide 2

Eight Steps to Forecasting 1. Determine use of forecast - what objective are we trying to obtain? 2. Select items or quantities to be forecasted. 3. Determine time horizon of forecast: • Short-range (less than three months) • Medium-range (3 months to 3 years) • Long-range (3+ years) 4. Select forecasting model or models. 5. Gather data needed to make forecast. 6. Validate forecasting model. 7. Make forecast. 8. Implement results. Slide 3

Types of Forecasts Slide 4

Qualitative Models • Qualitative models attempt to incorporate judgmental or subjective factors into forecasting model. • Opinions by experts, individual experiences and judgments, and other subjective factors may be considered. • Qualitative models are especially useful when subjective factors are expected to be very important or when accurate quantitative data are difficult to obtain. • Qualitative models are also useful for long-term forecasting. Slide 5

Time - Series and Causal (Associative) Models Time-series Models: • Time-series models attempt to predict future by using historical data. • Models make assumption that what happens in future is a function of what has happened in past. 2. Causal Models: • As with time-series models, causal models also rely on quantitative data. • Bivariate and multivariate regression models are examples of associative models. 1. Slide 6

Time series Components n The four Components of Time Series are: 1. 2. 3. 4. Trend Seasonal Cyclical Random Slide 7

General Forms of Time-series Models n There are two general forms of time-series models: • Most widely used is multiplicative model, which assumes forecasted value is product of four components. Forecast = (Trend). (Seasonality). (Cycles). ( Random) • Additive model adds components together to provide an estimate is also available. It is stated as: Forecast = Trend + Seasonality + Cycles + Random Slide 8

Moving Average n MA is a series of arithmetic means n Used if little or no trend n Used often for smoothing • Provides overall impression of data over time Equation: n MA = (Actual value in previous k periods) / k Slide 9

Weighted Moving Averages n Used when trend is present • Older data usually less important Weights based on intuition Equation: n k-period weighted moving average = n n (weight for period i) (actual value in period i) (weights) Slide 10

Exponential Smoothing n n n A form of weighted moving average • Weights decline exponentially • Most recent data weighted most Requires smoothing constant ( ) ranges from 0 to 1 is subjectively chosen Equation: Ft= Ft-1 + ( At-1 - Ft-1 ) Slide 11

Measuring Forecast Error n Mean Absolute Deviation (MAD). n MAD = |forecast error| / T = |At - Ft| / T n Select an which minimizes MAD. Slide 12

Causal Models n n Goal of causal forecasting model is to develop best statistical relationship between dependent variable and independent variables. Most common model used in practice is regression analysis. In causal forecasting models, when one tries to predict dependent variable usingle independent variable it is called a simple regression model. When one uses more than one independent variable to forecast dependent variable, it is called a multiple regression model. Slide 13

Monitoring and Controlling Forecasts Tracking Signal Ø Ø Measures how well the forecast is predicting actual values. Tracking Signal is the ratio of running sum of forecast errors (RSFE) to mean absolute deviation (MAD). Tracking Signal = RSFE/MAD Ø Ø Ø If forecasts are continually high or low, the forecast has a bias error. Good tracking signal has low values. For a forecast to be “in control” it is suggested to consider it in control up to + 3 MAD. Slide 14