二手车交易价格预测

数据探查

  1. 读取数据
df = pd.read_csv("./car_price.csv")
df.head(3)
  1. 数据信息
df.info()
  1. 空值查看
df.isnull().sum()
  1. 删除空值
df.dropna(inplace=True)

清洗处理

  1. 字段提取数字
    删除空格后面的字符串
df["mileage"] = df["mileage"].map(lambda x : float(x.split(" ")[0]))
df["mileage"].head(3)

df["engine"] = df["engine"].map(lambda x : float(x.split(" ")[0]))
df["engine"].head(3)

df["max_power"] = df["max_power"].map(lambda x : float(x.split(" ")[0]))
df["max_power"].head(3)

正则表达式提取字段中最大值

import re

def parse_rpm(torque):
    torque = torque.replace(",", "")
    return max([float(x) for x in re.findall("\d+", torque)])

df["torque"] = df["torque"].map(parse_rpm)
df["torque"].head(3)

统计分析

  1. 删除name字段
df.drop(columns=["name"], inplace=True)
df.head(3)
  1. object类统计
df.select_dtypes(include=["object"]).describe()
  1. float类统计
df.select_dtypes(include=["float"]).describe()
  1. 相关性矩阵
import seaborn as sns
import matplotlib.pyplot as plt
%matplotlib inline

plt.figure(figsize=(8,8))
sns.heatmap(df.corr(), annot=True)

可以看出销售价格（selling_price）和最大马力（max_power）强相关，其次是引擎容积（engine）、年份（year）

特征工程

  1. 类别字段one-hot
cat_columns = ["fuel", "seller_type", "transmission", "owner"]
from sklearn.preprocessing import OneHotEncoder
oneHotEncoder = OneHotEncoder(drop='first')

cat_features = oneHotEncoder.fit_transform(df[cat_columns]).toarray()
cat_features
  1. 数值字段标准化
num_columns = ["mileage", "engine", "max_power", "torque", "seats"]
from sklearn.preprocessing import StandardScaler

standardScaler = StandardScaler()
num_features = standardScaler.fit_transform(df[num_columns])
num_features
  1. 构建x和y
X = np.hstack([cat_features, num_features])
X

y = df["selling_price"].to_numpy()
y

模型训练

  1. 数据集划分
from sklearn.model_selection import train_test_split

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)
X_train.shape, X_test.shape, y_train.shape, y_test.shape

投票分类器 线性回归 SVM kNN回归 随机森林

Bagging

Adaboost回归

XGBoost

stacking
  1. 随机森林
from sklearn.ensemble import RandomForestRegressor
from sklearn.metrics import accuracy_score

random_model = RandomForestRegressor(n_estimators=300, random_state=42, n_jobs=-1)
random_model.fit(X_train, y_train)

y_pred = random_model.predict(X_test)
y_pred

random_model.score(X_train, y_train)
random_model.score(X_test, y_test)

交叉验证

网格搜索

模型评估

之前训练的随机森林模型与LR对比

  1. 训练线性回归模型
from sklearn.linear_model import LinearRegression

lr_model = LinearRegression()
lr_model.fit(X_train, y_train)
lr_model.score(X_train, y_train)
lr_model.score(X_test, y_test)

模型存储

  1. joblib.dump，存储random_model、oneHotEncoder、standardScaler
model_dir = "./flask-carprice/models"
import joblib

joblib.dump(random_model, f"{model_dir}/random_model.joblib")
joblib.dump(oneHotEncoder, f"{model_dir}/oneHotEncoder.joblib")
joblib.dump(standardScaler, f"{model_dir}/standardScaler.joblib")

在线预估

  1. 初始化flask
from flask import Flask, render_template, request
import pandas as pd
import numpy as np
import model

app = Flask("carprice_predict")


def get_predict_result(df):
    # datajson = """{"year":{"6046":2017},"selling_price":{"6046":2100000},"km_driven":{"6046":48000},"fuel":{"6046":"Diesel"},"seller_type":{"6046":"Individual"},"transmission":{"6046":"Automatic"},"owner":{"6046":"Second Owner"},"mileage":{"6046":17.9},"engine":{"6046":2143.0},"max_power":{"6046":136.0},"torque":{"6046":3000.0},"seats":{"6046":5.0}}"""
    # df = pd.read_json(datajson)

    cat_columns = ["fuel", "seller_type", "transmission", "owner"]
    load_cat_fatures = model.carPriceModel.onehot.transform(df[cat_columns]).toarray()

    num_columns = ["mileage", "engine", "max_power", "torque", "seats"]
    load_num_fatures = model.carPriceModel.scaler.transform(df[num_columns])

    load_final_fatures = np.hstack([load_cat_fatures, load_num_fatures])
    result = model.carPriceModel.predictor.predict(load_final_fatures)

    print(result)
    return result


@app.route("/predict", methods=["get", "post"])
def predict():
    result = None
    if request.method == "POST":
        data = dict(request.form)
        df = pd.DataFrame([data.values()], columns=data.keys())
        result = str(get_predict_result(df))
    return render_template("predict.html", result=result)


app.run(host="0.0.0.0", port=5010)
  1. 加载模型
import joblib


class CarPriceModel:
    def __init__(self):
        self.onehot = None
        self.scaler = None
        self.predictor = None

    def load_models(self):
        print("load_models")
        model_dir = "./models"
        self.onehot = joblib.load(f"{model_dir}/oneHotEncoder.joblib")
        self.scaler = joblib.load(f"{model_dir}/standardScaler.joblib")
        self.predictor = joblib.load(f"{model_dir}/random_model.joblib")


carPriceModel = CarPriceModel()
carPriceModel.load_models()
  1. html页面
<!DOCTYPE html>
<html lang="en">
<head>
    <meta charset="UTF-8">
    <title>二手车价格预估</title>
</head>
<body>

{% if result %}
<h1>预估结果：{{ result }}</h1>
{% endif %}

<form action="/predict" method="post">
    <table>
        <tr>
            <th>请输入fuel：</th>
            <td><input type="text" name="fuel"></td>
        </tr>
        <tr>
            <th>请输入seller_type：</th>
            <td><input type="text" name="seller_type"></td>
        </tr>
        <tr>
            <th>请输入transmission：</th>
            <td><input type="text" name="transmission"></td>
        </tr>
        <tr>
            <th>请输入owner：</th>
            <td><input type="text" name="owner"></td>
        </tr>
        <tr>
            <th>请输入mileage：</th>
            <td><input type="text" name="mileage"></td>
        </tr>
        <tr>
            <th>请输入engine：</th>
            <td><input type="text" name="engine"></td>
        </tr>
        <tr>
            <th>请输入max_power：</th>
            <td><input type="text" name="max_power"></td>
        </tr>
        <tr>
            <th>请输入torque：</th>
            <td><input type="text" name="torque"></td>
        </tr>
        <tr>
            <th>请输入seats：</th>
            <td><input type="text" name="seats"></td>
        </tr>
        <tr>
            <th>提交：</th>
            <td><input type="submit" name="submit" value="提交"></td>
        </tr>
    </table>
</form>

</body>
</html>

服务部署

  1. 将项目打包上传到linux服务器
tar zcvf flask-carprice.tar.gz flask-carprice
  1. 上传后解压
tar zxvf flask-carprice.tar.gz
  1. 安装依赖
pip3 install -i https://pypi.tuna.tsinghua.edu.cn/simple -r requirements.txt
  1. 启动命令
nohup python3 app.py 1>log.log 2>&1 &

  1. 配置安全组


  2. 访问页面
    公网ip:5010/predict


最后编辑于 ：
©著作权归作者所有,转载或内容合作请联系作者
平台声明：文章内容（如有图片或视频亦包括在内）由作者上传并发布，文章内容仅代表作者本人观点，简书系信息发布平台，仅提供信息存储服务。

推荐阅读更多精彩内容