New Classification Model for Fault Detection in 3W Dataset

[yantavares]

This PR was made by the team Black Oil Pyrates in HACKTUDO.

Description:

This PR introduces a new ClassificationModel class to the repository, which enables the training and evaluation of custom machine learning models to classify faults in the 3W dataset. The ClassificationModel allows for flexible model integration, supporting different algorithms for the identification and classification of class faults.

---

Code Overview:

• ClassificationModel class:
  • Attributes:
    • model: Allows the user to specify a custom classification model (e.g., Random Forest).
    • scaler: A StandardScaler instance used for feature normalization.
  • Methods:
    • load_and_prepare_data: Loads and preprocesses data from parquet files, handling timestamp cleaning and feature extraction.
    • train_model: Trains the provided machine learning model on the training dataset.
    • predict_single_file: Makes predictions on a single file, returning actual and predicted class labels along with the timestamps.
    • plot_predictions_with_timestamp: Visualizes the time-series predictions alongside actual class values using Plotly.
    • complete_analysis: Automates the entire process from data loading to prediction and visualization.
    • compare_models: Compares multiple models, training and testing them on different files, and evaluates their performance using accuracy and F1-score.

---

Example Usage:

from sklearn.ensemble import RandomForestClassifier
from classification import ClassificationModel

model = RandomForestClassifier(n_estimators=200, max_depth=12, random_state=42, class_weight='balanced')
dataset_path = './dataset/2'

# Initialize the classification model with Random Forest
classification_model = ClassificationModel(dataset_path=dataset_path, model=model)

# Get the list of parquet files
all_files = [os.path.join(dataset_path, f) for f in os.listdir(dataset_path) if f.endswith(".parquet")]

# Run the complete analysis
classification_model.complete_analysis(y_axis='T-TPT', files=all_files, train_size=0.8)

---

Dependencies:

• Pandas: For data loading, cleaning, and manipulation.
• Scikit-learn: For implementing various machine learning models and performance metrics.
• Plotly: For generating interactive visualizations of the time-series predictions.

This PR adds a flexible and scalable way to train machine learning models for fault classification in the 3W dataset, with support for custom models like Random Forest and interactive plotting capabilities.