Article directory
Use the Forecast Wizard to forecast values in a time series. The Forecast Wizard uses the Microsoft Time Series algorithm, which is a regression algorithm for forecasting continuous columns such as product sales.
Each predictive model must contain a sequence of cases, the column that distinguishes the different points in the sequence. For example, using historical data to forecast sales over several months, the dates are listed as case series.
Other Microsoft algorithms create models (such as decision tree models) that rely on given input columns to predict predictable columns. But time-series models make predictions based only on trends derived by the algorithm from the original dataset when the model was created.
1. Prediction based on box office data
Using 50 movie information as predictions:
Note: The SHOWTIME column needs to be sorted before making predictions
Here, the prediction accuracy will be lower due to the time interval of each movie.
2. Timestamp - convert text to numeric value
If the time parameter is not found in the timestamp when making predictions, the possible reason is that the data is not converted to a numeric type, but is stored in text mode.
The correct way is to turn time into numbers: just click Convert to Numbers in the image below.
3. Predicting the NBA's points per game
Take the following data as a prediction:
The data below is to convert more than 700 pieces of NBA data into a pivot table, and then average the points per game (pointsPg) to obtain the average of the points per game each year, and predict future data based on the known data. .
4. Create a mining structure
Create a predictive model with default steps.
If the KEY TIME is not specified when the mining structure is created, an error will be reported when the algorithm is used:
so use YEAR as the KEY TIME:
then follow the steps to create it.
5. Management mining structure
6. Add the model to the structure
After creating the predictive model, you need to add the model to the structure
- Select the timing algorithm.
The Microsoft Time Series algorithm uses ARIMA analysis in conjunction with decision tree-based linear regression to analyze time-dependent data, such as monthly sales data or annual profits. The patterns discovered by the algorithm can be used to predict values at future time steps . The algorithm can be customized to use the decision tree method or ARIMA, or both.
2. Select the logistic regression algorithm
The Microsoft logistic regression algorithm is a regression algorithm suitable for regression modeling. This algorithm is one of Microsoft's neural network algorithms and is obtained by eliminating hidden layers. The algorithm supports prediction of discrete and continuous attributes.
This time, the raw NBA data that has not been sorted out is selected:
7. Prediction based on COVID-19 data
Some of the data are as follows:
In addition, historical forecast information and deviations can be displayed:
