Modular python class to use Random Forest Classifier and make predictions without re-training data. Does search to find best suitable hyper parameters to the given dataset. Evaluates and saves the statistics, also logs every single action using a logging mechanism.

• Doesn't need to re-train data
• Don't need to manually preprocess data
• Predicts using the best hyper parameters
• Saves statistics
• Logs every single action
• Modular
• Easy to use

• data
  • data.csv # your data that you want to use to train the model
• log
  • data_unique_datetime_identifier.txt # simply a log file
• model
  • data_model_encoders.pickle # encoders that are used to encode the data in the preprocessing step
  • data_model_feature_names.pickle # feature names
  • data_model_value_name.pickle # output name
• statistics
  • data_model_datetime_identifier_confusion_matrix.png # confusion matrix
  • data_model_datetime_identifier_decision_tree.dot # decision tree of the first tree
  • data_model_datetime_identifier_statistics.txt # statistics like accuracy, precision, recall, f1-score, etc.

from forester import Forester

# Initialize Forester
## Assumes that the data is in the './data/data.csv' file and the default delimiter is ','
## When we set train=True, it will train the model and save the required files
forester = Forester(train=True)

# Create your prediction data
val = [0,...,'Example', 1]

# call make_prediction method
## It will return the prediction
prediction = forester.make_prediction(val)

# Print the prediction
print(prediction)

Example usages from different datasets can be found in the Example.py file.

• Python 3.6+
• Scikit-learn
• Pandas
• NumPy
• SciPy