Energy Prediction

Overview

Energy consumption patterns are vital for resource planning, optimizing utility management, and developing sustainable energy policies. Communities rely on accurate data about different energy sources, such as electricity and natural gas, to make informed decisions. However, predicting the type of energy being consumed based on regional, seasonal, and utility-based factors is challenging due to the diverse and complex interactions between these variables. By developing an effective classification model to predict the type of energy consumption, we can enhance energy planning efforts, improve infrastructure management, and promote efficient resource utilization across various communities.

Dataset

The dataset includes the following features:

• year
• data_class
• month
• value
• com_name
• com_type
• data_field
• com_county
• geometry_id
• full_fips
• unit
• uer_id
• data_stream
• utility_display_name
• number_of_accounts
• Georeference

Methodology

The following steps were taken to build the heart attack risk prediction model:

1. Data Preprocessing:

   • Handling missing values.
   • Feature scaling.
   • Encoding categorical variables.

2. Feature Engineering:

   • Selecting the top 10 important features using RandomForestClassifier.

3. Model Building:

   • Implementing several classification algorithms including LogisticRegression, svm, DecisionTree, RandomForest and aive_bayes
   • Hyperparameter tuning to optimize model performance.

4. Model Evaluation:

   • Evaluating models using metrics like accuracy, precision, recall, and F1-score.
   • Using confusion matrix to assess model performance.

Model Performance

The RandomForest model was optimized with the following parameters:

• n_estimators=100
• max_depth=30
• min_samples_split=2
• min_samples_leaf=1
• bootstrap'= False
• random_state=42

How to Use

To use the model to predict heart attack risk for new data, follow these steps:

1. Clone the repository:

   git clone https://github.com/safwankmohd/Energy-Prediction.git

2. Install the required packages:

   pip install -r requirements.txt

3. Load the trained model and scaler:

   import pickle
   import pandas as pd
   from sklearn.preprocessing import StandardScaler
   
   with open('Energy_prediction.pkl', 'rb') as file:
       rf1 = pickle.load(file)
   
   with open('Energy_prediction.pkl', 'rb') as file:
       scaler = pickle.load(file)

4. Create a DataFrame with the new data and make predictions:

   numerical_features = [ 'value', 'number_of_accounts']  # Adjust these based on your dataset
   
   unseen_data = pd.DataFrame({
   'value': [5000.0], 
   'com_name': ['Sodus'],  
   'com_type': ['Village'],  
   'data_field': ['all_other_(o)'], 
   'com_county': ['Wayne'],
   'unit': ['MWh'], 
   'number_of_accounts': [50], 
   'Georeference': ['POINT (-77.061462 43.236085)']
   })
   
   unseen_data_encoded = pd.get_dummies(unseen_data, columns=[ 'com_type', 'unit', ], drop_first=True)
   
   for col in ['com_name', 'Georeference', 'com_county', 'data_field']:
   freq_encoding = X[col].value_counts(normalize=True)  # Calculate the frequency of each category
   unseen_data_encoded[col] = X[col].map(freq_encoding)  # Map the original values to their frequency
   
   unseen_data_encoded = unseen_data_encoded.reindex(columns=X_train.columns, fill_value=0)
   
   scaler = StandardScaler()
   unseen_data_encoded[numerical_features] = scaler.fit_transform(unseen_data_encoded[numerical_features])
   predicted_class = loaded_model.predict(unseen_data_encoded)
   print("Predicted class:", predicted_class[0])

Conclusion

This project demonstrated a strong balance between training and test performance, showing robustness without significant overfitting. The model effectively captured the complex patterns in the dataset and maintained high accuracy across various classes. Despite the presence of missing data and potential class imbalance, the model performed well overall, making it a reliable choice for classification tasks in this dataset. Further improvements could be made by addressing limitations such as class imbalance and enhancing feature engineering.