awholeneworld/machinelearning

Manual

def auto_ml(dataset, model)

Select features randomly.
Run algorithms with the selected combination.

Parameter

dataset: DataFrame to be used
model: A model to be used

How to operate

1. Randomly select features from the list of numeric features.
2. Encoding and Scaling using scale_encode_combination().
3. Run the selected algorithm; one of [ test_kmeans(), test_gaussian(), test_clarans(), test_dbscan() and test_mean_shift() ].

Examples

df = pd.read_csv(‘housing.csv’)
df.fillna(df.mean(), inplace=True)
medianHouseValue = df['median_house_value']
df.drop(['median_house_value'], axis=1, inplace=True)
auto_ml(df, ‘kmeans’)

Return

All the results of the selected model.

def scale_encode_combination(dataset, numerical_feature_list, categorical_feature_list)

Scaling and Encoding with 15 combinations.

Parameters

dataset: DataFrame to be scaled and encoded
numerical_feature_list: Features to scale
categorical_feature_list: Features to encode

How to operate

1. for in scalers [StandardScaler(), MinMaxScaler(), RobustScaler(), MaxAbsScaler(), Normalizer()]
2. for in encoders [OrdinalEncoder(), OneHotEncoder(), LabelEncoder()]
3. Save each dataset in dictionary

Examples

for combination in feature_combination_list:
    data_combination = scale_encode_combination(dataset, combination, ['ocean_proximity'])
    for data_name, data in data_combination.items():
        data = data[combination]
        test_kmeans(data)
        test_gaussian(data)
        test_clarans(data)
        test_dbscan(data)
        test_mean_shift(data)

Return

Dictionary included all the dataframe combinations of Scalers and Encoders.