Linear Regression Examples (Simple)

These are just some toy examples to demonstrate write-once execute-many approach. They are provided so that you can tweak parameters in cfg.json of each example to build some more understanding of both SDK and working of regression algorithm.

Linear Regression model expects the data to be normally distributed. Meaning, histogram of the data takes a bell shaped curve.

Build

You can build the program once:

make build

Run

Copy the dataset to current working directory

make prb=problem1 copy

Execute the program:

make prb=problem1 run

Output

Program creates scratch folder under the current working directory. You should see following files after successfully running the program:

data

data folder contains test and training data created according to cfg.json. For instance linear/problem1 has following:

• test.csv
• training.csv

plot

plot folder contains various plots that were requested in cfg.json. For instance linear/problem1 has the following plots:

• hist-tv
• scatter-sales-tv

report

report folder stores two files test.json and train.json.

{
  "n": 40,
  "r2": -5.624073665997599,
  "mse": 0.27417851128875
}

where

• n: Number of samples
• r2: Coefficient of determination
• mse: Mean squared error

model.json

{
  "coeffs": {
    "tv": 0.5695823728693639
  },
  "intercept": 0.2158978690050924
}

where

• coeffs: Regression coefficients
• intercept: Constant