Ensemble Classifier

A Multiprocess Classifier Written Using Linux Pipes and FIFOs.

How to Build

This project is a cmake project. There is also a Makefile file that can be used on supported systems to run the cmake. The result will be in build/bin/ folder.

How to Run

The project can be run by this command:

./ensemble <validation_directory> <weights_directory>

Where <validation_directory> is the directory that holds the dataset.csv and labels.csv file and <weights_directory> is the directory that has the weights of each linear classifier.

Execution steps

1. The ensemble classifier with fork-exec linear classifier as many as there are classifier files in the <weights_directory>. It will pass the information that linear classifiers need through linux pipes.
2. Each linear classifier will do a linear classification on each of the 1000 data points inside dataset.csv file in the <validation_directory>. Each of them will write the result of their classifications inside a named pipe (created by using mkfifo(3) system call).
3. The ensemble classifier will also fork-exec a voter process which will read the classification results of each linear classifier after they are done and for each data point, choose the class that has the most votes (more linear classifiers have classified that data point with that class).
4. The voter process will write the results for each data point in another named pipe between the voter and ensemble classifier processes and the ensemble will calculate the accuracy of the classification by comparing the results to the actual classes found in labels.csv file inside <validation_directory> and output the results.