Linear Regression to predict a simple formula result using Weka

Question

I have a simple train data as follow. The result is calculated by formula (price * 100).

Train Data

Price, Result

1, 100

15, 1500

2.3, 230

5.23, 523

3.245, 324.5

100, 1000

Test Data

Price, Result

4.567, 456.7 (this result 456.7 to be predicted by formula price * 100)

Is there a way to do it by Weka?

Answer 1

I have fixed your datasets, turning them into valid CSV format:

Train

Price,Result
1,100
15,1500
2.3,230
5.23,523
3.245,324.5
100,10000

NB: The last line also required 10,000, not 1000 as Result value to comply with your rule of Result = Price * 100.

Test

Price,Result
4.567,456.7

Build model

You can build a model, either using weka.classifiers.functions.LinearRegression or weka.classifiers.functions.SimpleLinearRegression from the terminal like this (and output the generated rule):

java -cp weka.jar weka.classifiers.functions.SimpleLinearRegression -t train.csv -T test.csv

NB: You will need to adjust paths to the weka.jar and your train/test CSV files.

Predictions

If you want to output predictions for the test set, then you can use something like this:

java -cp weka.jar weka.classifiers.functions.SimpleLinearRegression -t train.csv -T test.csv -classifications weka.classifiers.evaluation.output.prediction.PlainText

NB: You will need to adjust paths to the weka.jar and your train/test CSV files.