Feature Selection with Hybrid TFFS

Description

The get_list_feature_tffs_hybrid function is used for feature selection based on high occurrence frequency during the Random Forest construction process, while also supporting integration with traditional feature selection methods.

Syntax

get_list_feature_tffs_hybrid(df, number_of_runs, n_estimators, percent, type=None, percent_hybrid=None)

Parameters

Valid Values for type

If type is used, one of the following feature selection methods can be chosen:

Usage

Using Random Forest Only for Feature Selection

import pandas as pd
from sff.app import get_list_feature_tffs_hybrid

# Sample DataFrame
data = {
    "Class": [0, 1, 0, 1, 0, 1, 0, 1, 0, 1],
    "Feature1": [5, 8, 6, 7, 5, 8, 6, 7, 5, 8],
    "Feature2": [1, 2, 3, 4, 5, 6, 7, 8, 9, 10],
    "Feature3": [10, 20, 10, 20, 10, 20, 10, 20, 10, 20]
}
df = pd.DataFrame(data)

# Run function
selected_features = get_list_feature_tffs_hybrid(df, number_of_runs=10, n_estimators=100, percent=20)
print("Selected Features:", selected_features)

Integrating with Mutual Information (MI)

selected_features = get_list_feature_tffs_hybrid(df, number_of_runs=10, n_estimators=100, percent=75, type="MI", percent_hybrid=50)
print("Selected Features:", selected_features)

Integrating with Recursive Feature Elimination (RFE)

selected_features = get_list_feature_tffs_hybrid(df, number_of_runs=15, n_estimators=200, percent=75, type="RC", percent_hybrid=50)
print("Selected Features:", selected_features)

Integrating with Mutual Information (MI) - Example Code

import pandas as pd
from sff.app import get_list_feature_tffs_hybrid

# Sample DataFrame
data = {
    "Class": [0, 1, 0, 1, 0, 1, 0, 1, 0, 1],
    "Feature1": [5, 8, 6, 7, 5, 8, 6, 7, 5, 8],
    "Feature2": [1, 2, 3, 4, 5, 6, 7, 8, 9, 10],
    "Feature3": [10, 20, 10, 20, 10, 20, 10, 20, 10, 20]
}
df = pd.DataFrame(data)

# Run function
selected_features = get_list_feature_tffs_hybrid(df, 5, 20, 75, "MI", 50)
print("Selected Features:", selected_features)

Notes

• If type is not provided, the function will use only Random Forest for feature selection.
• If type is provided, the function will integrate Random Forest with the specified traditional method for optimal feature selection.
• percent_hybrid is applicable only when type is used.
• Note: percent (4th parameter) must be greater than percent_hybrid.
• Note: df must have the class column as the first column.