Function evalute_csv

union/Santos/evaluation.py:91–133

(which_benchmark, which_method, topk)

def evalute_csv(which_benchmark, which_method, topk):
    if which_benchmark == 0:
        current_benchmark = 'webtable'
        groundtruth_path = '/data_ssd/webtable/small_query/ground_truth.csv'
    elif which_benchmark == 1:
        current_benchmark = 'opendata'
        groundtruth_path = '/data_ssd/opendata/small_query/ground_truth.csv'
    elif which_benchmark == 2:
        current_benchmark = 'webtable_large'
        groundtruth_path = '/data_ssd/webtable/large_query/ground_truth.csv'
    elif which_benchmark == 3:
        current_benchmark = 'opendata_large'
        groundtruth_path = '/data/opendata/large_query/ground_truth.csv'
    
    if which_method == 0:
        current_method = 'santos'
    else:
        current_method = 'tus'
    
    #result_path = 'result/webtable_benchmark_true_result_by_santos_full_20.csv'
    result_path = 'result1/' + current_benchmark + '_' + current_method + '_top' + str(topk) + '.csv'
    df_gt = pd.read_csv(groundtruth_path)
    df_result = pd.read_csv(result_path)
    column_data = df_result["query_table"]
    unique_values = column_data.unique()
    #对结果里的每个表计算精度，召回率，以及map
    precisions = []
    recalls = []
    save_dict = {}
    #保存一个字典，key是table name，value是list精度召回，存成json
    for table in unique_values:
        precision, recall, len_intersection, len_result, len_gt = calculate_metrics(table, df_result, df_gt)#计算这一个表的精度召回
        precisions.append(precision)
        recalls.append(recall)
        save_dict[table] = (precision ,recall, len_intersection ,len_result, len_gt)
    
    
    precision = np.mean(precisions)
    recall = np.mean(recalls)

    save_dict_to_json(save_dict, 'result/webtable_evaluation_5.json')
    print("Precision:", precision)
    print("Recall:", recall)


def evalute_dict():