BUAA研究生机器学习作业二手车价格预测.zip

import pandas as pd import numpy as np import seaborn as sns from lightgbm import LGBMRegressor from matplotlib import pyplot as plt from sklearn.ensemble import BaggingRegressor from sklearn.ensemble import GradientBoostingRegressor from sklearn.ensemble import AdaBoostRegressor from sklearn.tree import ExtraTreeRegressor from sklearn.tree import DecisionTreeRegressor from sklearn.ensemble import RandomForestRegressor import xgboost as xgb from xgboost import XGBRegressor train_pd = pd.read_csv("train.csv", sep=' ') test_pd = pd.read_csv("test.csv", sep=' ') continuous_variable = ['power', 'kilometer', 'price', 'v_0', 'v_1', 'v_2', 'v_3', 'v_4', 'v_5', 'v_6', 'v_7', 'v_8', 'v_9', 'v_10', 'v_11', 'v_12', 'v_13', 'v_14'] for i in continuous_variable: fig = sns.distplot(np.log1p(train_pd[i])) fig_save = fig.get_figure() fig_save.savefig('{}.png'.format(i), dpi=300) fig_save.clear() corr = train_pd[continuous_variable].corr() ax = plt.subplots(figsize=(20, 16)) # 调整画布大小 fig = sns.heatmap(corr, vmax=.8, square=True, annot=True) # 画热力图 annot=True 表示显示系数 fig_save = fig.get_figure() fig_save.savefig('heatmap.png', dpi=300) fig_save.clear() # 设置刻度字体大小 plt.xticks(fontsize=20) plt.yticks(fontsize=20) df = pd.concat([train_pd, test_pd], axis=0) df = df.replace('-', np.nan) print(df.isnull().sum()) df['fuelType'] = df['fuelType'].fillna(0).astype(np.float64) df['gearbox'] = df['gearbox'].fillna(0).astype(np.float64) df['bodyType'] = df['bodyType'].fillna(0).astype(np.float64) df['model'] = df['model'].fillna(0).astype(np.float64) df['notRepairedDamage'] = df['notRepairedDamage'].fillna(0).astype(np.float64) df['creatDate_year'] = df['creatDate'].apply(lambda x: str(x)[:4]).astype(np.int32) df['creatData_month'] = df['creatDate'].apply(lambda x: str(x)[4:6]).astype(np.int32) df['creatDate_day'] = df['creatDate'].apply(lambda x: str(x)[6:8]).astype(np.int32) df['regDate_year'] = df['regDate'].apply(lambda x: str(x)[:4]).astype(np.int32) df['regData_month'] = df['regDate'].apply(lambda x: str(x)[4:6]).astype(np.int32) df['regDate_day'] = df['regDate'].apply(lambda x: str(x)[6:8]).astype(np.int32) df['power'] = df['power'].apply(lambda x: 600 if x > 600 else x) bin = [i*10 for i in range(31)] df['power_bin'] = pd.cut(df['power'], bin, labels=False) bin = [i*10 for i in range(24)] df['model_bin'] = pd.cut(df['model'], bin, labels=False) print(df['power_bin']) v_high_correspondence = ['v_0', 'v_3', 'v_8', 'v_12'] for i in v_high_correspondence: for j in v_high_correspondence: df[i + "+" + j] = df[i] + df[j] for i in v_high_correspondence: for j in v_high_correspondence: df[i + '-' + j] = df[i] - df[j] for i in v_high_correspondence: for j in v_high_correspondence: df[i + '*' + j] = df[i] * df[j] y_id = test_pd["SaleID"] # 获取测试集的id，以便后面保存为csv文件 df.drop(['SaleID'], axis=1, inplace=True) df.drop(['regDate'], axis=1, inplace=True) df.drop(['creatDate'], axis=1, inplace=True) df.drop(['regionCode'], axis=1, inplace=True) df.drop(['name'], axis=1, inplace=True) train_pd = df[df['price'].notnull()] test_pd = df[df['price'].isnull()] # model = RandomForestRegressor() model = LGBMRegressor( n_estimators=300000, learning_rate=0.02, boosting_type='gbdt', objective='regression_l1', max_depth=-1, num_leaves=31, min_child_samples=20, feature_fraction=0.8, bagging_freq=1, bagging_fraction=0.8, lambda_l2=2, random_state=2022, metric='mae', ) # model = RandomForestRegressor(n_estimators=25, random_state=42, oob_score=True) model.fit(train_pd.drop('price', axis=1), np.log1p(train_pd['price'])) y_pred = model.predict(test_pd.drop('price', axis=1)) y_pred = np.expm1(y_pred) result = pd.DataFrame({'SaleID': y_id, 'price': y_pred.astype(np.int32)}) result.to_csv("submission.csv", index=False)