阿里AI天池大赛-二手车交易价格预测-回归算法与模型融合

1：报名地址

https://tianchi.aliyun.com/competition/entrance/231784/introduction?spm=5176.12281925.0.0.1a3c7137nPlTRm

 2：排名分数

3：模型源码

 废话不多说，直接上源码

 

## 基础工具
import numpy as np
import pandas as pd
import warnings
import matplotlib
import matplotlib.pyplot as plt
import seaborn as sns
from scipy.special import jn
from IPython.display import display, clear_output
import time

warnings.filterwarnings('ignore')
%matplotlib inline

## 模型预测的
from sklearn import linear_model
from sklearn import preprocessing
from sklearn.svm import SVR
from sklearn.ensemble import RandomForestRegressor,GradientBoostingRegressor

## 数据降维处理的
from sklearn.decomposition import PCA,FastICA,FactorAnalysis,SparsePCA

import lightgbm as lgb
import xgboost as xgb

## 参数搜索和评价的
from sklearn.model_selection import GridSearchCV,cross_val_score,StratifiedKFold,train_test_split
from sklearn.metrics import mean_squared_error, mean_absolute_error

## 通过Pandas对于数据进行读取 (pandas是一个很友好的数据读取函数库)
Train_data = pd.read_csv('./used_car_train_20200313.csv', sep=' ')
TestA_data = pd.read_csv('./used_car_testB_20200421.csv', sep=' ')

## 输出数据的大小信息
print('Train data shape:',Train_data.shape)
print('TestA data shape:',TestA_data.shape)


## 通过.head() 简要浏览读取数据的形式
Train_data.head()


## 通过 .info() 简要可以看到对应一些数据列名，以及NAN缺失信息
Train_data.info()

## 通过 .columns 查看列名
Train_data.columns


TestA_data.info()

## 通过 .describe() 可以查看数值特征列的一些统计信息
Train_data.describe()

TestA_data.describe()


Train_data[['regDate','creatDate']]


Train_data['regDate_year'] = Train_data['regDate'].astype('str').apply(lambda x: int(x[0:4]))
Train_data['regDate_month'] =Train_data['regDate'].astype('str').apply(lambda x: int(x[4:6]))
Train_data['regDate_day'] =Train_data['regDate'].astype('str').apply(lambda x: int(x[6:]))

Train_data['creatDate_year'] = Train_data['creatDate'].astype('str').apply(lambda x: int(x[0:4]))
Train_data['creatDate_month'] =Train_data['creatDate'].astype('str').apply(lambda x: int(x[4:6]))
Train_data['creatDate_day'] =Train_data['creatDate'].astype('str').apply(lambda x: int(x[6:]))

TestA_data['regDate_year'] = TestA_data['regDate'].astype('str').apply(lambda x: int(x[0:4]))
TestA_data['regDate_month'] =TestA_data['regDate'].astype('str').apply(lambda x: int(x[4:6]))
TestA_data['regDate_day'] =TestA_data['regDate'].astype('str').apply(lambda x: int(x[6:]))

TestA_data['creatDate_year'] = TestA_data['creatDate'].astype('str').apply(lambda x: int(x[0:4]))
TestA_data['creatDate_month'] =TestA_data['creatDate'].astype('str').apply(lambda x: int(x[4:6]))
TestA_data['creatDate_day'] =TestA_data['creatDate'].astype('str').apply(lambda x: int(x[6:]))

numerical_cols = Train_data.select_dtypes(exclude = 'object').columns
print(numerical_cols)

categorical_cols = Train_data.select_dtypes(include = 'object').columns
print(categorical_cols)

## 选择特征列
feature_cols = [col for col in numerical_cols if col not in ['SaleID','price','regDate','creatDate']]
feature_cols = [col for col in feature_cols if 'Type' not in col]

## 提前特征列，标签列构造训练样本和测试样本
X_data = Train_data[feature_cols]
Y_data = Train_data['price']

X_test  = TestA_data[feature_cols]

print('X train shape:',X_data.shape)
print('X test shape:',X_test.shape)

X_data.columns

## 定义了一个统计函数，方便后续信息统计
def Sta_inf(data):
    print('_min',np.min(data))
    print('_max:',np.max(data))
    print('_mean',np.mean(data))
    print('_ptp',np.ptp(data))
    print('_std',np.std(data))
    print('_var',np.var(data))
    
    
print('Sta of label:')
Sta_inf(Y_data)

## 绘制标签的统计图，查看标签分布
plt.hist(Y_data)
plt.show()
plt.close()

X_data = X_data.fillna(-1)
X_test = X_test.fillna(-1)

## xgb-Model
xgr = xgb.XGBRegressor(n_estimators=1600, learning_rate=0.05, gamma=0, subsample=0.8,\
        colsample_bytree=0.8, max_depth=12) #,objective ='reg:squarederror'

scores_train = []
scores = []

## 10折交叉验证方式
sk=StratifiedKFold(n_splits=10,shuffle=True,random_state=0)
for train_ind,val_ind in sk.split(X_data,Y_data):
    
    train_x=X_data.iloc[train_ind].values
    train_y=Y_data.iloc[train_ind]
    val_x=X_data.iloc[val_ind].values
    val_y=Y_data.iloc[val_ind]
    
    xgr.fit(train_x,train_y)
    pred_train_xgb=xgr.predict(train_x)
    pred_xgb=xgr.predict(val_x)
    
    score_train = mean_absolute_error(train_y,pred_train_xgb)
    scores_train.append(score_train)
    score = mean_absolute_error(val_y,pred_xgb)
    scores.append(score)

print('Train mae:',np.mean(score_train))
print('Val mae',np.mean(scores))

def build_model_xgb(x_train,y_train):
    model = xgb.XGBRegressor(n_estimators=1600, learning_rate=0.05, gamma=0, subsample=0.8,\
        colsample_bytree=0.8, max_depth=12) #, objective ='reg:squarederror'
    model.fit(x_train, y_train)
    return model

def build_model_lgb(x_train,y_train):
    estimator = lgb.LGBMRegressor(num_leaves=127,n_estimators = 1600)
    param_grid = {
        'learning_rate': [0.01, 0.05, 0.1, 0.2],
    }
    gbm = GridSearchCV(estimator, param_grid)
    gbm.fit(x_train, y_train)
    return gbm

## Split data with val
x_train,x_val,y_train,y_val = train_test_split(X_data,Y_data,test_size=0.3)

print('Train lgb...')
model_lgb = build_model_lgb(x_train,y_train)
val_lgb = model_lgb.predict(x_val)
MAE_lgb = mean_absolute_error(y_val,val_lgb)
print('MAE of val with lgb:',MAE_lgb)

print('Predict lgb...')
model_lgb_pre = build_model_lgb(X_data,Y_data)
subA_lgb = model_lgb_pre.predict(X_test)
print('Sta of Predict lgb:')
Sta_inf(subA_lgb)

print('Train xgb...')
model_xgb = build_model_xgb(x_train,y_train)
val_xgb = model_xgb.predict(x_val)
MAE_xgb = mean_absolute_error(y_val,val_xgb)
print('MAE of val with xgb:',MAE_xgb)

print('Predict xgb...')
model_xgb_pre = build_model_xgb(X_data,Y_data)
subA_xgb = model_xgb_pre.predict(X_test)
print('Sta of Predict xgb:')
Sta_inf(subA_xgb)

## 这里我们采取了简单的加权融合的方式
val_Weighted = (1-MAE_lgb/(MAE_xgb+MAE_lgb))*val_lgb+(1-MAE_xgb/(MAE_xgb+MAE_lgb))*val_xgb
val_Weighted[val_Weighted<0]=10 # 由于我们发现预测的最小值有负数，而真实情况下，price为负是不存在的，由此我们进行对应的后修正
print('MAE of val with Weighted ensemble:',mean_absolute_error(y_val,val_Weighted))

sub_Weighted = (1-MAE_lgb/(MAE_xgb+MAE_lgb))*subA_lgb+(1-MAE_xgb/(MAE_xgb+MAE_lgb))*subA_xgb

## 查看预测值的统计进行
plt.hist(Y_data)
plt.show()
plt.close()

sub = pd.DataFrame()
sub['SaleID'] = TestA_data.SaleID
sub['price'] = sub_Weighted
sub.to_csv("./baseline_lgb_xgb.csv",index=False)

4：提分要领     

        1：时间类型数据(regDate, creatDate)

        2：模型融合，使用XGBoost,LightGBM

        3：增加新的特征，对brand进行统计 brand_amount, brand_price_max, brand_price_min, brand_price_median, brand_price_sum, brand_price_std, brand_price_mean         

        4：使用XGBoost，超参数设置

        5：模型融合：回归任务中的加权融合 根据各个模型的最终预测表现分配不同的权重，比如准确率高的模型给予更高的权重，准确率低的模型给予较小的权重

5：证书展示

 

6：相关知识补充

1：回归分析算法

        研究因变量（目标）与自变量（特征）之间的关系

        帮助数据科学家，更好的选择最佳的变量集，用于建立预测模型

        使用场景：价格预测，数量预测

        

        

 2：常见的回归算法

        线性回归 / 逻辑回归

        多项式回归 Polynomial Regression 多项式是一种常用的特征构造方法

        岭回归 Ridge Regression

        套索回归 Lasso Regression

        弹性回归 ElasticNet Regression

        岭回归和套索回归的混合技术，同时使用L2和L1正则

        

3：其他回归算法

        分类用于回归：SVM，KNN，CART

        集成学习：RF，GBDT, XGBoost, LightGBM