Data mining by example forecasting and cross prediction using microsoft time series
Download as PPTX, PDF2 likes1,510 views
This document discusses using Microsoft Time Series for forecasting and cross prediction. It describes setting up a data mining project in SQL Server to create a time series mining structure and model using sales history from AdventureWorksDW. The model is processed, viewed, and queried to forecast trends and make predictions. Additional data can be added to the model using DMX, and cross prediction is demonstrated by creating new queries and mining structures. Microsoft Time Series provides an intuitive way to perform time series forecasting and analysis in SQL Server data mining projects.
Data mining by example forecasting and cross prediction using microsoft time series
- 1. Data Mining By Example – Forecasting and Cross Prediction Using Microsoft Time Series by Shaoli Lu
- 2. Microsoft Time Series • Microsoft Time Series algorithm provides a unique approach to time series forecasting that is both intuitive and accurate. It is used to forecast future series points based on past history
- 3. Goal • To forecast and cross-predict based on past sales history
- 4. Prerequisite • An SQL Server instance created (2005 or above) • SQL Server Analysis Service (SSAS) – Multidimensional Feature Installed (this is used to host and browse the mining structures; cube is not required for data mining!) • AdventureWorksDW database attached (download from CodePlex - tailor to the SQL Server version you have) • Visual Studio 2010 or above with SQL Server Data Tools (SSDT) installed
- 5. My Demo Setup • Visual Studio 2010 • SQL Server 2012
- 6. Create Data Mining Project • Name the project as DM Forecasting (DM = Data Mining)
- 8. Create Data Source and Impersonation
- 11. Create Data Source View
- 14. Create Mining Structure • Use relational data source • Choose Microsoft Time Series model • Select Data Source View • Select key, input and predict • Name the mining structure and model • Add Amount to the mining structure and change it to Predict • Tune Algorithm Parameters
- 23. Deploy the mining structure and model
- 24. Process the mining model • This is also called “training the model”
- 27. Mining Model Viewer • Forecast trend • Select items to predict • View standard deviation • View forecasting mode in a tree graph • View regression formula • Use Generic Content Tree View to inspect stats details
- 31. Mining Model Prediction • Convert to DMX query • Adjust the DMX query by adding FLATTENED clause • Add filters to the DMX query • View query results
- 38. Adding Additional Data • DMX EXTEND_MODEL_CASES clause will add additional data to the existing data
- 42. Cross Prediction • Create new named query AllRegions for cross- prediction • Create a query for a specific region and model • Create a new mining structure for cross- prediction • Deploy and process the cross-prediction mining structure and model • Use REPLACE_MODEL_CASES in the DMX query for cross-prediction
- 50. Browse mining model on SQL Server • Time Series Chart View • Model Tree Graph • Prediction Query Window • DMX • Query result
- 56. Summary • Microsoft Time Series is a powerful data mining model, yet it is intuitive to build, train and use • It is useful in forecasting future event points • Algorithm Parameters can be tuned • DMX EXTEND_MODEL_CASES clause will add additional data to the existing data • Use REPLACE_MODEL_CASES in the DMX query for cross-prediction • Relational database can be used for data mining; cube is not required
- 57. The End