Regression

In this sample program, we will introduce how to use the linear regression function jubaregression through the Jubatus Client.

By using linear regression, we can estimate the output from the input data. This is useful for the power consumption forecast, stock price prediction, and etc.

Abstract of sample program

In this sample, we will use a rent program to estimate the rent of a house, from the condition of the distance from the station, space, age, etc.

At first, please download the rent-data, from (rent-data.csv). It is used to training the regression model in jubaregression .

Next, housing data for the estimation are read from an external file (of YAML format). jubaregression will predict the rent by using the model trained, and return the predicted value for client.

For example, once client input a housing data of [(distance from station: 15min); (space: 32 m^2); (age: 15 years)], the estimated rent 80,000 JPY will be returned.

Processing flow

The flow of development using Jubatus Client is following:

  1. Connection settings to jubaregression
    Setting the HOST and RPC port of jubaregression .
  2. Prepare the training data
    Get the rent-data from the downloaded CSV file.
  3. Data training (update the model)
    Use the train method to send the rent-data to jubaregression and training the model there.
  4. Prepare the data for estimation
    Pre-process and send the estimation data to jubaregression .
  5. Predict the data
    Use the estimate method to predict the input data (step.4) by using the model trained at step.3.
  6. Return the result
    Output the return value of estimate as the results.

Sample Program

Currently, we have no sample programs except Python. (We welcome your contribution!)