algebra_misconceptions / create_exp2_results.py
nanote's picture
Upload 12 files
8880178 verified
raw
history blame
1.03 kB
import pandas as pd
output_file_path = 'outputs/experiment_2_gpt-4-turbo_100iters.csv'
outputs_df = pd.read_csv(output_file_path)
print(outputs_df.head())
def calculate_precision_recall(df):
# Calculate precision and recall
true_positives = df['Correct'].sum()
total_predicted = len(df)
total_actual = df['Misconception ID'].nunique()
precision = true_positives / total_predicted if total_predicted else 0
recall = true_positives / total_actual if total_actual else 0
return precision, recall
# Overall precision and recall
overall_precision, overall_recall = calculate_precision_recall(outputs_df)
# Precision and recall per topic
topic_precision_recall = outputs_df.groupby('Topic').apply(calculate_precision_recall).apply(pd.Series)
topic_precision_recall.columns = ['Precision', 'Recall']
# Display results
print(f"Overall Precision: {overall_precision:.3f}")
print(f"Overall Recall: {overall_recall:.3f}")
print("\nPrecision and Recall per Topic:")
print(topic_precision_recall)