Knowt
Note
Studied by 11 peopleNoteStudied by 52 peopleNoteStudied by 48 peopleNoteStudied by 7 peopleNoteStudied by 10 peopleNoteStudied by 153 people
Forecasting
Page 1: Forecasting Models
Page 2: Introduction to Forecasting
Definition of Forecasting: Predicting future events based on available data (time series or other).
Importance:
Influences decision-making in present operations.
Applications in Operations:
Forecasting demand for products and services.
Assessing manpower needs.
Planning inventory and material requirements.
Identifying risks for remediation (e.g., Nokia vs. Ericsson).
Page 3: Characteristics of Forecasts
Uncertainty: Forecasts typically involve a degree of uncertainty.
Detail in Predictions: A good forecast is a range rather than a single figure.
Should include mean value and standard deviation.
Accuracy metrics: high and low value ranges.
Precision in Aggregated Forecasts: Generally more accurate.
Accuracy Over Time: Accuracy decreases with longer time horizons.
Page 4: What Makes a Good Forecast?
Timeliness: Should be generated in a timely manner.
Accuracy: Must aim for high accuracy.
Reliability: Consistency in results.
Meaningful Units: Must be presented in units relevant to the context.
Clarity: Should be documented and easy to understand.
Simplicity: Should be user-friendly.
Page 5: Forecast Horizons in Operation Planning
Product Family Sales: Short-term forecasts for products.
Labor Needs: Planning based on shifts and manpower requirements.
Resource Requirements: For short and long-term needs.
Capacity Planning: Aligning capacities with long-term sales patterns and trends.
Page 6: Subjective Forecasting Methods
Sales Force Composites: Estimates from sales personnel.
Customer Surveys: Gathering feedback directly from customers.
Jury of Executive Opinion: Expert insights aggregated into a consensus.
The Delphi Method: Involves multiple rounds of anonymous expert opinions to reach consensus.
Page 8: Objective Forecasting Methods
Primary Methods:
Causal Models:
Representation: Y (forecasted quantity) is influenced by multiple variables (X1, X2, ... Xn).
Formulation: Y = f(X1, X2, ... Xn).
Typical relationship: Linear regression model.
Time Series Methods: Focus on historical data for trend analysis.
Page 9: Demand Patterns (Time Series Data)
Purely Random: No identifiable trend.
Increasing Pattern: Rising over time.
Curvilinear Patterns: Trends that are non-linear.
Seasonal Trends: Recognizable increases/decreases tied to specific periods.
Page 10: Evaluation of Forecasts
Forecast Error Calculation: For a period, forecast error εt = Ft - Dt.
Forecasting Accuracy Metrics:
MAE (Mean Absolute Error) = (1/n) Σ |εi|
MSE (Mean Squared Error) = (1/n) Σ(εi)²
RMSE (Root Mean Squared Error) = √(1/n) Σ(εi)².
Aim for random distribution of forecast errors.
Page 11: Moving Averages
Concept: Arithmetic average of the most recent 'n' observations.
Page 12: Moving Average Example
3-month MA: (Oct + Nov + Dec)/3 = 258.33
6-month MA: (Jul + Aug + Dec)/6 = 249.33
12-month MA: (Jan + Feb + Dec)/12 = 205.33.
Data Table:
Demand Month: Data provided for each month from January to December.
Page 13: Summary of Moving Averages
Advantages:
Simplicity in computation.
Stability in forecasts.
Disadvantages:
Requires substantial historical data.
Tends to lag trends.
Complexity in data relationships may be overlooked.
Page 14: Moving Average Lags an Increasing Trend
Graph indicating how moving averages lag behind actual trends.
Page 15: Example-1: Time Series Forecasting
Data Context: Weekly sales over 8 weeks.
Tasks: Calculate 2- and 4-period moving average forecasts.
Determine least Mean Absolute Deviation (MAD).
Forecast for the subsequent week.
Page 16: Example-1 Solution
Calculated Forecasts:
F3 =(32 + 34)/2 = 33
F5 = (32 + 34 + 35 + 33)/4 = 33.5
MAD Calculations:
MAD for 2-period = 1.42;
MAD for 4-period = 1.38 (least MAD).
Page 17: Naive Method of Forecasting
Simple Method: Uses most recent past data.
Example Context: Given demand data and forecasts for June.
Page 18: Naive Method Calculations
Data Presented: Actual and Forecast Demand calculations for several months.
Page 19: Mean-Simple Average Method
Overview of how to calculate average demand over years.
Page 20: Simple Moving Average Method
3-Year Period Calculation: Use past years for forecasting the demand of the 7th year.
Page 21: Simple Moving Average Method Continuation
Error Summary: Summary of errors associated with forecasts.
MAD Calculation: MAD calculated as approximately 54.
Page 22: Error Metrics and Interpretation
Displays actual and forecast error values.
Mean Absolute Percentage Error (MAPE): Measures forecasting accuracy as a percentage.
Page 23: Breakdown of MAPE
Additional Error Metrics:
Detailed MAPE calculations for various months and forecast values.
Page 24: Detailed MAPE Results
Continued metrics calculations.
Page 25: Weighted Moving Averages
Link to video resource for additional learning on weighted moving averages.
Page 26: Exponential Smoothing Methods
Concept: Assigns exponentially decreasing weights to observations based on recency.
Smoothing Parameters: Parameter 'alpha' determines the weightings.
Page 27: Exponential Smoothing Abstract Data
Context: Given data up to the 6th year to forecast the 7th year.
Page 28: Practical Application of Exponential Smoothing
Forecasting demand using exponential smoothing with a defined α.
Page 29: Equation for Exponential Smoothing
Formula representation: Ft+1 = Ft + α(At-1 - Ft).
Page 30: Working Exponential Smoothing Examples
Data Context: Utilizes actual and forecast demand with α calculations.
Page 31: Exponential Smoothing Computation Demonstration
Detailed Walkthrough: Explaining the calculations using given α.
Page 32: Another Example of Exponential Smoothing
Resource Link: Video referencing exponential smoothing methodologies.
Page 33: Additional Resource on Exponential Smoothing
Resource Link: Video in Hindi regarding the techniques in demand forecasting.
Page 34: Effect of α values on Forecasts
Stability Impact: How varying alpha values influence forecast variability.
Page 35: Additional Resource Link
Video about forecasting concepts.
Page 36: Linear Trend Line Concept
Linear Regression in Forecasting: Establishing relationships between demand and influencing factors.
Equation Overview: y = a + bx.
Page 37: Data Trend Analysis for Hi Tek Computer Services
Graphical representation of trend analysis based on historical data.
Page 38: Least Squares Calculations Overview
Computation Table Provided: A detailed look at necessary computations for forecasting.
Page 39: Visual Representation of Forecast Data
Graph Comparison: Shows actual demand versus linear trend line forecasts.
Page 40: Seasonal Adjustments
Definition of Seasonal Patterns: Regular fluctuations in demand over specific periods.
Page 41: Seasonal Factors in Forecasting
Methodology: Using seasonal factors to adjust baseline forecasts.
Page 42: Demand Data for Wishbone Farms
Case Study: Analysis of turkey demand over the years.
Page 43: Seasonal Factors for Wishbone Farms
Forecast Calculation Steps: Calculating seasonal adjustments based on historical data.
Page 44: Final Seasonal Demand Calculation
Final Adjustments: Seasonal adjustments applied based on forecasted demand.
Page 45: Linear Regression Methodology
Summary of linear regression as a forecasting tool.
Page 46: General Structure of Linear Regression
Explanation of the regression formula, encompassing demand relationships.
Page 47: Attendance Forecasting Case Study
Example Application: Developing a budget based on attendance forecasts related to winning streaks.
Page 48: Least Squares Computation Summary
Summary of computations determining the linear equation for predicting attendance.
Page 49: Practice Problem Overview
Context for Students: Engaging in real-world forecasting scenarios.
Page 50: Exponential Smoothing Example Overview
Demonstration of Exponential Smoothing Applications: Practical engagement with different values of α.
Page 51: Exponential Smoothing Data Representation
Detailing results for various smoothing values and their respective forecasts.
Page 52: Absolute Error and MAD Comparisons
Assessment of α Performance: Identifying the optimal α based on lowest MAD.
Page 53: MA vs ES Comparison
Summary of Differences: Outlines the distinctions between Moving Average and Exponential Smoothing methodologies.
NoteStudied by 11 peopleNoteStudied by 52 peopleNoteStudied by 48 peopleNoteStudied by 7 peopleNoteStudied by 10 peopleNoteStudied by 153 people