Skip to content

Latest commit

 

History

History
52 lines (33 loc) · 3.16 KB

README.md

File metadata and controls

52 lines (33 loc) · 3.16 KB
Raw

KNN and Decision Trees Analysis

This repository contains Python scripts for performing analysis using k-nearest neighbors (KNN) and decision trees algorithms. These algorithms are implemented using various libraries such as scikit-learn and Faiss.

Overview

The repository consists of the following files:

  • knn.py: Contains the implementation of the KNNClassifier class, which utilizes the Faiss library for efficient k-nearest neighbors classification.
  • decision_trees.py: Includes functions for training decision tree classifiers and analyzing their performance.
  • data.csv, train.csv, validation.csv, test.csv, AD_test.csv: These CSV files contain data used for training and testing the machine learning models. Each row represents a sample, with features such as longitude, latitude, and country statistics.

Files Description

knn.py

The knn.py file contains the implementation of the KNNClassifier class, which allows for k-nearest neighbors classification. Below are the key components of this file:

  • KNNClassifier Class: This class implements a k-nearest neighbors classifier using the Faiss library for efficient distance computations. It includes methods for fitting the model to training data (fit), predicting class labels for new data (predict), and calculating kNN distances (knn_distance).

decision_trees.py

The decision_trees.py file provides functions for training decision tree classifiers and evaluating their performance. Key functionalities of this file include:

  • Decision Tree Demo: This function demonstrates the usage of decision tree classifiers using scikit-learn's DecisionTreeClassifier. It includes training the classifier, making predictions, and computing accuracy.
  • Random Forest and XGBoost Loading: This file also includes functions for loading random forest and XGBoost classifiers using scikit-learn's RandomForestClassifier and XGBClassifier.
  • Plot Decision Boundaries: A utility function plot_decision_boundaries is provided for visualizing decision boundaries of classifiers.
  • Anomaly Detection: Another function anomaly_detection_demo is available for performing anomaly detection using k-nearest neighbors.
  • Train Decision Trees: This function trains decision trees with different hyperparameters and evaluates their performance on test and validation sets.
  • KNN Analysis: Lastly, there is a function KNN for analyzing the performance of the KNN classifier on different distance metrics and values of k.

Data Files Description

data.csv, train.csv, validation.csv, test.csv, AD_test.csv

These CSV files contain data used for training and testing the machine learning models. Each row represents a sample, with features such as longitude, latitude, and country statistics.

Usage

To utilize the functionalities provided in these files, follow the steps below:

  1. Clone the repository to your local machine.
  2. Ensure that you have the required dependencies installed (numpy, pandas, matplotlib, scikit-learn, Faiss).
  3. Run the desired Python script (e.g., python decision_trees.py) to execute the analysis.

Dependencies

  • NumPy
  • pandas
  • Matplotlib
  • scikit-learn
  • Faiss