MLPK Framework

Machine Learning PCA K-Fold Framework

Tutorial

1. Create venv with python3 -m venv v-env
2. Active your environment source v-env/bin/active
3. install all dependencies pip3 install -r requirements.txt
4. Set the mandatory parameters for the choosen algorithm parameters in the run.py
5. python3 run.py

Mandatory Parameters

• n_split: K of K-fold
• n_components: PCA number of components
• test_size: percentage of data used in the test part (0<test_size<1)
• start_date: date in in string format to add in dataset
• data_path: dataset path
• columns: columns of dataset (must have a y column)
• epochs: number of epochs

Algorithms Parameters

Neural Networks

algo_conf = {
    'name': 'NN',
    'n_layers': 3,
    'input_dim': 2,
    'n_neurons': [5, 8, 1],
    'list_act_func': ['tanh', 'tanh', 'sigmoid'],
    'loss': 'mean_squared_error',
    'optimizer': 'sgd',
    'metric': 'accuracy'
}

Naive Bayes Binary Cls

algo_conf = {
    'name': 'naive_bayes_bcls',
    'priors': None,
    'var_smoothing': 0.1,
    'metric': 'accuracy'
}

Logistic Regression Binary Cls

algo_conf = {
    'name': 'logistic_regression_bcls',
    'solver': 'liblinear',
    'metric': 'accuracy'
}

SVM Binary Cls

algo_conf = {
    'name': 'svm',
    'metric': 'accuracy'
}

Decision Tree Binary Cls

algo_conf = {
    'name': 'decision_tree',
    'metric': 'accuracy'
}

Random Forest Binary Cls

algo_conf = {
    'name': 'random_forest',
    'metric': 'accuracy'
}

XGBoot Binary Cls

algo_conf = {
    'objective': 'binary:logistic',
    'name': 'xgboost',
    'metric': 'accuracy'
}

LightGBM Binary Cls

algo_conf = {
    'name': 'lightgbm',
    'metric': 'accuracy'
}    

Sample data in ./data

It's just a sample dataset, you can adapt your own dataset.

• Bitcoin: 2013-08-19 to 2016-07-19
  • Data market price from charts.bitcoin

References

• TensorFlow 2
• scikit-learn
• Machine Learning Mastery
• Neural Network on Keras
• PCA on Keras
• Adam Optimization
• Google Trends Normalization code