Interesting code snippets

import pandas as pd
from sklearn.base import BaseEstimator, TransformerMixin

class ImportantAttributeSelector(BaseEstimator, TransformerMixin):
# This Transformer helpes in selecting top attributes having linear correlation and eliminating a few

def __init__(self, target_attribute):
    self.target_attribute = target_attribute
    pass

def fit(self, X, y=None):
    return self

def transform(self, X, attr_count):

    # calculate correlation matrix
    corr_matrix = X.corr()

    # project the target attribute
    target_corr_matrix = corr_matrix[self.target_attribute]

    # target_corr_matrix represents linear relationship in the range of [-1, 1]
    # coefficients close to 0, are having less to no linear correlation

    # convert all the values to absolute and sort in ascending order
    tr_target_corr_matrix = list(map(lambda item: (abs(item[1]), item[0]), target_corr_matrix.items()))
    tr_target_corr_matrix.sort(reverse=True)

    # create the list of attributes and return the required attribute
    attr_set = list(map(lambda item: item[1], tr_target_corr_matrix))
    return attr_set[:attr_count]

numeric_data = data.select_dtypes(exclude=["object"]).copy()