main.py

# using python 3.8
# 311148902
# 311300784

import pandas as pd
import Adaboost


# extract data from files
# for each dataset:
#   run adaboost 100 times
#       split dataset to train and test
#       run adaboost on the test set and get 8 best rules and their weights
#       do some calculations on the 8 best rules:
#           create a list
#           for k in range(8):
#               compute the empirical error of the function Hk on the test set
#               compute the true error of Hk on the training set
#

def main():
    pd.set_option('display.max_columns', None)

    """## Data preparation"""
    # importing body_temperature dataset
    body_temperature = pd.read_table("HC_Body_Temperature.txt", sep='    |   ', names=["temp", "gender", "heart_rate"],
                                     engine='python', encoding="ISO-8859-1")

    # reformat gender column: female = -1, male = 1
    body_temperature.gender = body_temperature.gender.apply(lambda x: -1 if x == 1 else 1)
    body_temperature_y = body_temperature.gender
    body_temperature_x = body_temperature.drop(labels="gender", axis=1)

    # importing iris dataset
    iris = pd.read_csv("iris.data", names=["sepal_length", "sepal_width", "petal_length", "petal_width", "iris_type"],
                       encoding="ISO-8859-1", low_memory=False)

    # dropping the undesired data from table. we only interested in 2nd and 3rd columns as features
    iris = iris.drop(labels=["sepal_length", "petal_width"], axis=1)

    # ignore Iris-setosa rows
    iris = iris.loc[iris.iris_type != "Iris-setosa"]

    # reformat iris_type column: Iris-versicolor = -1, Iris-virginica = 1
    iris.iris_type = iris.iris_type.apply(lambda x: -1 if x == "Iris-versicolor" else 1)
    iris.head()
    # all datasets are currently pandas.dataframe

    # splitting iris to features and labels
    iris_y = iris.iris_type
    # print(iris_y.shape)
    iris_x = iris.drop(labels="iris_type", axis=1)

    # single adaboost run.
    # input: features panda dataframe, labels panda dataframe
    # output: list of 8 best rules with errors and weights

    print("######## Iris results ########")
    sum_of_8_best_complicated_rules_stats = \
        [{"empirical_error_on_test": 0.0, "true_error_on_training": 0.0} for _ in range(8)]
    for i in range(100):
        curr_run = Adaboost.run(iris_x, iris_y)
        for index, rule in enumerate(sum_of_8_best_complicated_rules_stats):
            rule["empirical_error_on_test"] += curr_run[index]["empirical_error_on_test"]
            rule["true_error_on_training"] += curr_run[index]["true_error_on_training"]

    avg_errors_of_8_best = [{"empirical_error_on_test": 0.0, "true_error_on_training": 0.0} for _ in range(8)]
    for index, rule in enumerate(avg_errors_of_8_best):
        rule["empirical_error_on_test"] = sum_of_8_best_complicated_rules_stats[index]["empirical_error_on_test"] / 100
        rule["true_error_on_training"] = sum_of_8_best_complicated_rules_stats[index]["true_error_on_training"] / 100
        print("Rule #", index, "info:\n"
              , "\tempirical_error_on_test -", rule["empirical_error_on_test"]
              , "\ttrue_error_on_training -", rule["true_error_on_training"])

    print("\n######## Body temperture results ########")
    sum_of_8_best_complicated_rules_stats2 = \
        [{"empirical_error_on_test": 0.0, "true_error_on_training": 0.0} for _ in range(8)]
    for i in range(100):
        curr_run2 = Adaboost.run(body_temperature_x, body_temperature_y)
        for index, rule in enumerate(sum_of_8_best_complicated_rules_stats2):
            rule["empirical_error_on_test"] += curr_run2[index]["empirical_error_on_test"]
            rule["true_error_on_training"] += curr_run2[index]["true_error_on_training"]

    avg_errors_of_8_best2 = [{"empirical_error_on_test": 0.0, "true_error_on_training": 0.0} for _ in range(8)]
    for index, rule in enumerate(avg_errors_of_8_best2):
        rule["empirical_error_on_test"] = sum_of_8_best_complicated_rules_stats2[index]["empirical_error_on_test"] / 100
        rule["true_error_on_training"] = sum_of_8_best_complicated_rules_stats2[index]["true_error_on_training"] / 100
        print("Rule #", index, "info:\n"
              , "\tempirical_error_on_test -", rule["empirical_error_on_test"]
              , "\ttrue_error_on_training -", rule["true_error_on_training"])


if __name__ == "__main__":
    main()