190 likes | 474 Views
CID: A Software Tool for Intermittent Demand Analysis. CELDi Spring 2010 Meeting Manuel D. Rossetti, PhD. P. E. Intermittent Demand. Examples: spare parts, “slow moving” products Characteristics High probability of zero demand Zero demands clustered together
E N D
CID: A Software Tool for Intermittent Demand Analysis CELDi Spring 2010 Meeting Manuel D. Rossetti, PhD. P. E.
Intermittent Demand • Examples: spare parts, “slow moving” products • Characteristics • High probability of zero demand • Zero demands clustered together • May also have high variability in order size • Active area of research • Standard techniques tend to overestimate after demand occurs • Specialized techniques have been developed and analyzed CELDi Intermittent Demand Forecaster
Forecasting Software JMP Minitab ($1195) R OpenForecast SAS ForecastPro ($1295) CID Forecaster CELDi Intermittent Demand Forecaster
The Software • Purpose: aid managers in choosing a forecasting method and stocking policy for a particular item that has intermittent demand • Implemented in Java • Main functionality based on a sub-package of the Java Simulation Library (Rossetti, 2008) • Object-oriented • Code re-use • Observer pattern CELDi Intermittent Demand Forecaster
Software Framework CELDi Intermittent Demand Forecaster
Software Capability CELDi Intermittent Demand Forecaster
Software Functions- Intermittent Statistics • Time series statistics on: • All demands • Non-zero demands • Time between non-zero demands • Interval between zero demands • State transition probabilities • States= {zero, non-zero} CELDi Intermittent Demand Forecaster
Software Functions- Forecasting • Forecasting • Markov Chain Autoregressive to All (MCARTA) technique (Varghese and Rossetti, 2009) • make a forecast • compare techniques • performance metrics • batch mode • Metaforecasting • classifier-based forecast technique selection CELDi Intermittent Demand Forecaster
Software Functions- Performance Metrics • Based on classification of Hyndman and Koehler (2005) Forecast Metric Error Metric Percentage Error Metric Relative Error Metric Scaled Error Metric Symmetric Percentage Error Metric Unbiased Percentage Error Metric Unbiased Mean Absolute Percentage Error (Collopy and Armstrong 2000) Mean Square Error CELDi Intermittent Demand Forecaster
Use Case: From Data to Stocking Policy CELDi Intermittent Demand Forecaster
Comparing Forecast Techniques • Prediction intervals • Forecast Variance • Plot Forecast • Table of Residuals • Statistics • Error • Intermittent CELDi Intermittent Demand Forecaster
Inventory Analysis • Step 1: Set demand during lead time distribution • 1a: Set manually • 2a: Estimate via forecasting • Step 2: Choose model • Step 3: Set model parameters • Step 4: Run Model 1a 1b 2 3 4 CELDi Intermittent Demand Forecaster
Inventory Results CELDi Intermittent Demand Forecaster
Summary • Purpose: aid managers in choosing a forecasting method and stocking policy for a particular item that has intermittent demand • Object-oriented design allows improvement to the forecasting package to be done efficiently • Capabilities • Data analysis • Forecasting (comparison, batch mode, etc.) • Inventory analysis • Also, allows simulation modeling of inventory system under intermittent demand CELDi Intermittent Demand Forecaster
Questions? CELDi Intermittent Demand Forecaster
Inventory Capability • Assume demand during lead time distribution (e.g., Gamma) • Analytical model • Demand mean estimated by the next period forecast • Demand variance proportional to Mean Absolute Deviation (MAD) • Moment match to find parameters for demand during lead time distribution • Simulation model • Demand generator: array of data, text file, database, stochastic process (e.g., Batch on/off process) CELDi Intermittent Demand Forecaster
Simulation Model 1 • Step 1: Choose dataset • Step 2: Choose demand generator • Step 3: Choose model • Step 4: Set parameters • 4a: Cost • 4b: Simulation • Step 5: Run Model 2 3 4a 4b 5 CELDi Intermittent Demand Forecaster
Simulation Results CELDi Intermittent Demand Forecaster