/Naive-Bayes-Random-Forest

Primary LanguagePython

Naive Bayes Random Forest Classifier

The project is a Random Forest from-scratch implementation, which relies on a Naive Bayes classifiers instead of decision trees.

Project covers:

  • Random Forest implementation
  • Naive Bayes implemenatation
  • Comparison between sklearn Random Forest and own implementation

More information was reported in the polish language here.

Installation

  1. Install poetry: https://python-poetry.org/docs/#installation
  2. Create an environment with poetry install
  3. Run poetry shell
  4. To run unit tests for your service use poetry run pytest or simply pytest within poetry shell.

Usage

import os
import sys
nb_dir = os.path.split(os.getcwd())[0]
if nb_dir not in sys.path:
    sys.path.append(nb_dir)

Import dependencies

import pandas as pd
from sklearn.preprocessing import OneHotEncoder
import numpy as np
from Project.NaiveBayesClassifier import NaiveBayes
from Project.BayesianRandomForest import RandomForest
from sklearn.model_selection import train_test_split

Read data

data = pd.read_csv('../data/agaricus-lepiota.data', header=None).to_numpy()
y, X = np.split(data, [1], axis=1)
X_train, X_test, y_train, y_test = train_test_split(X,y)

Preprocess

y_list = y_train.reshape(-1)
enc = OneHotEncoder()
X_train_enc = enc.fit_transform(X_train)
X_train_enc = X_train_enc.toarray()
X_test_enc = enc.transform(X_test)
X_test_enc = X_test_enc.toarray()

Fit and predict

rf_classifier = RandomForest(100, feature_bagging=False)
rf_classifier.fit(X_train_enc, y_list)
result = rf_classifier.predict(X_test_enc)

License

MIT