Random_Forest_Predict_Missing_Values.py

# -*- coding: utf-8 -*-
"""
Created on Fri Dec 27 20:58:48 2024

@author: ramio
"""

import pandas as pd
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import train_test_split
from sklearn.metrics import classification_report, confusion_matrix, accuracy_score,ConfusionMatrixDisplay
import matplotlib.pyplot as plt
from scipy.sparse import hstack
import nltk
from nltk.corpus import stopwords
import pickle

# NLTK Portuguese stopwords (only needed once)
nltk.download('stopwords')

# Load Portuguese stopwords
portuguese_stopwords = stopwords.words('portuguese')

# Load the dataset
file_path = 'Registo dos livros (Guardado automaticamente).xlsx'
df = pd.read_excel(file_path, header=1)

# Data Cleaning (drop column)
df.columns = df.columns.str.strip() 
df = df.drop(['Unnamed: 14'], axis=1)

#Filtering data (train and missing)
missing_data= df [df["Tema & Localização"].isna()] # Rows where 'Tema & Localização' is missing (missing_data)
train_data = df [df["Tema & Localização"].notna()] # Rows where 'Tema & Localização' is not missing (train_data)

# Calculating class counts
class_counts = train_data['Tema & Localização'].value_counts()
print(class_counts)

# Identifying rare classes (less than 5 samples)
rare_classes = class_counts[class_counts < 5].index
print(rare_classes)

# Replacing rare classes with a new label
train_data['Tema & Localização'] = train_data['Tema & Localização'].replace(rare_classes, 'Other')

#Features selection
x= train_data[['Titulo','Autor','Editora','Tema & Localização']]  
y= train_data['Tema & Localização']

# Converting text columns to numerical using TF-IDF
tfidf = TfidfVectorizer(stop_words=portuguese_stopwords, max_features=1000)

# Vectorizing each text column separately
x_tfidf_titulo = tfidf.fit_transform(x['Titulo'].fillna(''))  # Transform 'Titulo' column
x_tfidf_autor = tfidf.transform(x['Autor'].fillna(''))    # Transform 'Autor' column
x_tfidf_editora = tfidf.transform(x['Editora'].fillna(''))  # Transform 'Editora' column
x_tfidf_tema = tfidf.transform(x['Tema & Localização'].fillna(''))  # Transform 'Tema & Localização' column

# Combining the TF-IDF features from all columns into one feature matrix
x_combined = hstack([x_tfidf_titulo, x_tfidf_autor, x_tfidf_editora, x_tfidf_tema])

#Data split
x_train,x_test,y_train,y_test = train_test_split (x_combined,y, test_size=0.2, random_state=42)

#Train model
rf_model = RandomForestClassifier(n_estimators=100, random_state=42)
rf_model.fit(x_train,y_train)

# Making prediction on the test set
y_pred = rf_model.predict(x_test)

# Calculating and print accuracy
accuracy = accuracy_score(y_test, y_pred)
print(f'Accuracy: {accuracy * 100:.2f}%')

# Confusion matrix
print('confusion matrix:')
# Confusion matrix
ConfusionMatrixDisplay.from_predictions(
    y_test, y_pred, 
    cmap='Blues', 
    colorbar=True
)
plt.xticks(rotation=45, fontsize=5,ha='right')  
plt.yticks(fontsize=5)
plt.title('Confusion Matrix')
plt.show()

# Classification report for more evaluation metrics
print('Classification Report:')
print(classification_report(y_test, y_pred))


""""Predicting missing values"""

# Predict the missing values in 'Tema & Localização'
x_missing = missing_data[['Titulo', 'Autor', 'Editora','Tema & Localização']]  # Select features for rows with missing 'Tema & Localização'

# Vectorizing the missing data
x_missing_tfidf_titulo = tfidf.transform(x_missing['Titulo'].fillna(''))
x_missing_tfidf_autor = tfidf.transform(x_missing['Autor'].fillna(''))
x_missing_tfidf_editora = tfidf.transform(x_missing['Editora'].fillna(''))
x_missing_tfidf_tema = tfidf.transform(x_missing['Tema & Localização'].fillna(''))  # Transform 'Tema & Localização' column

# Combining the TF-IDF features for the missing data
x_missing_combined = hstack([x_missing_tfidf_titulo, x_missing_tfidf_autor, x_missing_tfidf_editora,x_missing_tfidf_tema])

# Predicting missing values for 'Tema & Localização'
y_missing_pred = rf_model.predict(x_missing_combined)

# Replaceing the missing values in the original dataframe with the predicted values
df.loc[df["Tema & Localização"].isna(), 'Tema & Localização'] = y_missing_pred

# Displaying the dataframe with the predicted values filled in
print(df.head())


# Saving the trained model
with open('book_category_model.pkl', 'wb') as f:
    pickle.dump(rf_model, f)