import pandas as pd

e = pd.read_csv('election.csv')

def find_best_split_col(df):
    feats = df.columns[:-1]
    target = df.columns[-1]
    rights = pd.Series([])
    for feat in feats:
        print("Considering splitting on {}...".format(feat))
        vals = df[feat].value_counts().keys()
        right = 0
        for val in vals:
            right += df[df[feat] == val][target].value_counts().max()
            print("  {}-{}:{}".format(feat,val,
                df[df[feat] == val][target].value_counts().max()))
        print("Splitting on {} gives us {} right.".format(feat,right))
        rights[feat] = right
    return rights

root = find_best_split_col(e).idxmax()
print("Root: split on {}!".format(root))


for val in e[root].value_counts().keys():
    print()
    branch = e[e[root] == val].drop(root, axis=1)
    node = find_best_split_col(branch).idxmax()
    print("  {}={}: split on {}!".format(root,val,node))