把數據集隨機切分為訓練集和測試集
method 1:
df = pd.read_csv('data/tgnb_merge.csv', encoding='utf-8') df.drop_duplicates(keep='first', inplace=True) # 去重，只保留第一次出現的樣本 df_test = df.sample(frac=0.1) df_train = pd.concat([df, df_test], axis=0) # 拼接 df_train.drop_duplicates(keep=False, inplace=True) # 去除所有重復的樣本 print df.shape, df_test.shape, df_train.shape # (3045, 12) (305, 12) (2740, 12)
method 2（推薦）:
df = pd.read_csv('data/tgnb_merge.csv', encoding='utf-8') # df.drop_duplicates(keep='first', inplace=True) # 去重，只保留第一次出現的樣本 df = df.sample(frac=1.0) # 全部打亂 cut_idx = int(round(0.1 * df.shape[0])) df_test, df_train = df.iloc[:cut_idx], df.iloc[cut_idx:] print df.shape, df_test.shape, df_train.shape # (3184, 12) (318, 12) (2866, 12)