calc.py

import os
import re

import pandas as pd
import yaml

query_csv = "mushra.csv"
query_df = pd.read_csv(query_csv)
query_df["index"] = query_df.index
print(query_df)

results_csv = "results.csv"
result_df = pd.read_csv(results_csv)

bad_file = "badguys.yaml"
with open("badguys.yaml", "r") as file:
    bad_worker_ids = file.read().split()

# Drop bad worker results
result_df = result_df[~result_df["WorkerId"].isin(bad_worker_ids)]

# Drop other stuffs
pattern = re.compile(r"^\d+_\d+_rating$")
rating_columns = [col for col in result_df.columns if pattern.match(col)]
result_df = result_df[rating_columns]

result_df = result_df.transpose()
result_df.index = result_df.index.astype(str).str.strip()
result_df.index = result_df.index.str.replace("_rating", "")


result_df["idx"] = result_df.index.map(lambda x: int(x.split("_")[0]))
result_df["querynum"] = result_df.index.map(lambda x: int(x.split("_")[1]))
result_df = result_df.sort_values(by=["idx", "querynum"])
# print(result_df)


whatever = [0, 0, 0]


def calculate_mushra(index, querynum):
    filtered_df = result_df[
        (result_df["idx"] == index) & (result_df["querynum"] == querynum)
    ]

    if filtered_df.empty:
        print(f"No data found for index: {index} and querynum: {querynum}")
        return
    mean_score = filtered_df.mean(axis=1)
    mean_score = mean_score.round(2)
    min_score = 0  # Anchor (minimum rating)
    max_score = 100  # Reference (maximum rating)

    normalized_scores = (mean_score - min_score) / (max_score - min_score) * 100

    print(f"{query_df.loc[index, 'subtype']}, querynum {mean_score}")
    return normalized_scores.values[0]


for j in range(26, 42, 1):
    for i in [1, 2, 3]:
        whatever[i - 1] += calculate_mushra(j, i)
    print("----")


print(whatever[0] / 16)
print(whatever[1] / 16)
print(whatever[2] / 16)