Regression of Used Car Prices
osama zaman
Regression of Used Car Prices
Regression of Used Car Prices
Language
Python
Competition Notebook
Regression of Used Car Prices
Private Score 65375.40529
Best Score 65375.40529 V2
Input
import pandas as pd
import re
from sklearn.preprocessing import LabelEncoder
import xgboost as xgb
Load training data
data = pd.read_csv("/kaggle/input/playground-series-s4e9/train.csv")
Handling missing values
data["fuel_type"] = data["fuel_type"].fillna("Unknown")
data["accident"] = data["accident"].fillna("Unknown")
data["clean_title"] = data["clean_title"].fillna("Unknown")
Feature Engineering
data["car_age"] = 2025 - data["model_year"]
data.drop("model_year", axis=1, inplace=True)
Extract Horsepower and Engine Size from 'engine' column
data["horsepower"] = data["engine"].apply(lambda x: float(re.search(r"(\d+(.\d+)?)HP", str(x)).group(1)) if pd.notna(x) and re.search(r"(\d+(.\d+)?)HP", str(x)) else None)
data["engine_size"] = data["engine"].apply(lambda x: float(re.search(r"(\d+(.\d+)?)L", str(x)).group(1)) if pd.notna(x) and re.search(r"(\d+(.\d+)?)L", str(x)) else None)
data.drop("engine", axis=1, inplace=True)
Convert categorical variables to binary
data["accident"] = data["accident"].apply(lambda x: 0 if x == "None reported" else 1)
data["clean_title"] = data["clean_title"].apply(lambda x: 1 if x == "Yes" else 0)
Encoding Categorical Variables
categorical_cols = ["brand", "model", "fuel_type", "transmission", "ext_col", "int_col"]
label_encoders = {}
for col in categorical_cols:
le = LabelEncoder()
data[col] = le.fit_transform(data[col])
label_encoders[col] = le # Store encoder for test set
One-Hot Encoding for low-cardinality categorical features
data = pd.get_dummies(data, columns=["fuel_type", "transmission", "ext_col", "int_col"], drop_first=True)
Save cleaned train data
train_df = data.copy()
Load competition test dataset
test_df = pd.read_csv("/kaggle/input/playground-series-s4e9/test.csv")
Apply same feature engineering as train_df
test_df["car_age"] = 2025 - test_df["model_year"]
test_df.drop("model_year", axis=1, inplace=True)
Extract numerical features from 'engine'
test_df["horsepower"] = test_df["engine"].apply(lambda x: float(re.search(r"(\d+(.\d+)?)HP", str(x)).group(1)) if pd.notna(x) and re.search(r"(\d+(.\d+)?)HP", str(x)) else None)
test_df["engine_size"] = test_df["engine"].apply(lambda x: float(re.search(r"(\d+(.\d+)?)L", str(x)).group(1)) if pd.notna(x) and re.search(r"(\d+(.\d+)?)L", str(x)) else None)
test_df.drop("engine", axis=1, inplace=True)
Convert categorical variables to binary in test_df
test_df["accident"] = test_df["accident"].apply(lambda x: 0 if x == "None reported" else 1)
test_df["clean_title"] = test_df["clean_title"].apply(lambda x: 1 if x == "Yes" else 0)
Apply Label Encoding to test_df safely
for col in categorical_cols:
if col in test_df.columns: # Ensure column exists before transformation
test_df[col] = test_df[col].apply(lambda x: label_encoders[col].classes_[0] if x not in label_encoders[col].classes_ else x)
test_df[col] = label_encoders[col].transform(test_df[col])
One-Hot Encoding for test data (ensuring same structure as train_df)
test_df = pd.get_dummies(test_df, columns=["fuel_type", "transmission", "ext_col", "int_col"], drop_first=True)
Ensure test set has the same feature columns as train set
missing_cols = set(train_df.columns) - set(test_df.columns)
for col in missing_cols:
test_df[col] = 0 # Add missing columns with default 0
Ensure column order matches between train and test
X_train = train_df.drop(columns=["id", "price"])
y_train = train_df["price"]
X_test = test_df.drop(columns=["id"], errors="ignore")[X_train.columns] # Align columns
Train XGBoost Regressor
xgb_model = xgb.XGBRegressor(
n_estimators=500,
learning_rate=0.05,
max_depth=7,
subsample=0.8,
colsample_bytree=0.8,
random_state=42
)
xgb_model.fit(X_train, y_train)
test_df["price"] = xgb_model.predict(X_test)
Prepare submission file
submission = test_df[["id", "price"]]
submission.to_csv("/kaggle/working/submission.csv", index=False)
print("XGBoost Model Training Completed. Submission file saved as 'submission.csv'.")
License
This Notebook has been released under the Apache 2.0 open source license.
Like this project
Posted May 11, 2026
A competition project for predicting used car prices using machine learning techniques.
Likes
0
Views
0
