Simple code in C to perform Linear Regression using Gradient Descent. The code allows the user to specify which variables to consider in the regression as well as defining the learning rate and number of iterations. This is a work in progress.
- Generate indicator variables for ordinal and categorical variables
- Add support for reading regression instructions from .txt file
- Add possibility of running multiple regression specifications at once
- Give the possibility of imputting missing values instead of completely eliminating them
- Add support for reading other file types
- NaiveMatrices.c : The C file with all the code for the necessary linear algebra operations to perform LinearRegression
- NaiveMatrices.h : Header file to make the functions and Matrix struct usable in the execution code
- Execution.c : The C file that runs the code to execute the linear regression and interact with the user
- data.csv : Weather forecasting dataset from this Kaggle competition
First compile the code by opening up your terminal and using:
gcc -o LinearReg NaiveMatrices.c Execution.cThen you can run it by executing:
./LinearRegThe table below is an example output of the code:
| Independent Variable | Parameter Value | Confidence Interval | p-value |
|---|---|---|---|
| purchase_amount | 2.9228 | (2.9002, 2.9454) | 0.0000 |
| age | -0.5634 | (-0.5763, -0.5505) | 0.0000 |
| loyalty_score | -1.3759 | (-2.5559, -0.1960) | 0.0223 |
So far, the code only works with numerical variables. It also only does a simple regression and does not consider the possibility of interaction or squared terms.