CAR PRICE PREDICTION

Riya Kant1, Yuvarj Aggarwal1, Tushar Gupta1, Shilpa Gupta2

1Student, Department of Computer Science JEMTEC, G r Noida

2Assistant Professor, Department of Computer Science JEMTEC, G r Noida

Abstract— Accurate prediction of automobile prices is important for good decision-making by automobile market participants and business competition. This study presents a comparison of three popular regression methods (linear regression, random forest regression, and decision tree regression) used for traffic cost estimation. The model is trained and analyzed using a comprehensive database containing various features such as car make, model, year, mileage, engine size and other irrelevant factors. Thanks to detailed data analysis, model design and preliminary procedures, the data becomes ready for modeling. Our regression models are then implemented and improved using cross-validation techniques to improve their performance. Statistical measures such as mean error (MAE), mean square error (MSE), and R-squared were used to evaluate the prediction accuracy of each model. The results show that random forest regression outperforms linear regression and decision tree regression in terms of prediction accuracy. Random forest regression shows excellent performance in handling non-linearities, interactions between features, and outliers present in the dataset. Its conditions allow the decision tree to reduce its bounds, resulting in good predictions. Linear regression, although simple and interpretable, often performs poorly when faced with relationships between features and target variables. Although decision tree regression is capable of capturing interactions, it can suffer from overfitting and poor generalization. This study provides useful information on the advantages and limitations of different methods for estimating traffic costs. It provides practical advice to automotive industry participants on selecting appropriate regression models for accurate and reliable vehicle price prediction. In summary, this study enables the estimation of vehicle costs by comparing regression methods. Using these insights, stakeholders can make informed decisions that will ultimately improve pricing strategies and market competitiveness in the automotive industry.

Keywords—car price prediction, regression analysis, Linear Regression, Random Forest Regression, Decision Tree Regression, predictive accuracy, feature engineering, model evaluation, automotive industry, market competitiveness.

1. INTRODUCTION

In today's automobile market, needs are changing, customers' preferences are changing and competition remains attractive, forecasting automobile prices is important, alerting all stakeholders, from individual buyers to large companies. The ability to accurately estimate costs is critical and helps determine pricing strategies, inventory management, and overall market competitiveness. In this context, regression analysis is important in modeling and predicting traffic prices according to many influencing factors.

The motivation for this study is the deep inconsistency in the automobile market and the increasing need for accurate price prediction. With the growth of online car sales platforms and the abundance of big data with many features, there is a significant opportunity to leverage advanced analytics to improve the accuracy of car price prediction. The overall goal of this project is to determine the best method to reliably estimate vehicle prices by comparing different models.

The basis of this study is to examine and compare the prediction performance of three commonly used methods in traffic cost prediction: linear regression, random forest regression and decision tree regression. Through careful analysis and evaluation, we aim to see which model is more accurate and powerful in predicting car prices in different situations. Moreover, in addition to identifying the most effective models, we are happy to offer participants in the automotive industry practical ideas and recommendations to improve their ownership at eight prices and increase their competitiveness. By better understanding the advantages and limitations of each regression method, we strive to help our partners make decisions that will lead to success, precision and stability in the dynamic auto market.

2. RELATED WORK

According to author Noor and Jan [1], They use multiple linear regression to construct a car price forecasting model. The data set was generated over a period of two months and included the following characteristics: size, cubic capacity, exterior color, date of publication of ad, number of ad views, power steering, mileage, transmission type, engine type, region, registered region, layout, edition, make and model year. With the results set, the researchers were able to achieve 98% predictability. The authors

proposed a prediction model based on a single machine learning algorithm in the relevant research mentioned above. However, it is noteworthy that the standard approach to machine learning algorithms did not produce impressive predictive results and could be improved by combining multiple machine learning methods into an ensemble.

According to author Gonggie [2], He proposed a model that would be developed using ANN (Artificial Neural Networks) to estimate the price of a used vehicle. He considered several attributes: mileage, estimated life of the car and make. The new model was developed to cope with non-linear data interactions, which was not the case with previous models using standard linear regression techniques. The nonlinear model was able to predict car prices better than other linear models with greater accuracy.

According to author Wu et al. [3], He performed an analysis of car price estimation using a knowledge-based neurofuzzy method. They took into account the following characteristics: model, year of manufacture and engine size. Their projection model had comparable findings to the simplified regression model. They created a special program called ODAV (Optimal Distribution of Auction Vehicles) because there is a strong demand for car dealers to deliver vehicles at the end of the lease year. This method offers information on the best car rates and also the place to get the best quality. A neighbor-centered k-nearest machine learning algorithm was used to predict the car's speed. This program appears to have been remarkably effective, having replaced more than two million vehicles.

According to author Putra, P. H., Azanuddin, A., Purba, B., & Dalimunthe, Y. A. [4], Currently in the era of cars that use renewable fuels such as electric cars that are highly supported by the government to impact on used cars based on these issues, an analysis is needed. Determining whether or not the price of buying or selling a used car is reasonable is one of the hurdles that the community faces when making a decision to buy or sell a car or vehicle. Therefore, most people choose the alternative of buying a used car, which is still good and usable. One way to predict the price is to use a machine learning method. In this study, they used random forest and decision tree methods to predict car prices. Research results of car price prediction analysis using random forest and decision tree methods have different percentage results. Using the random forest method, the accuracy is: 72.13%, while using the decision tree method, the accuracy is: 67.21%. Thus, it can be concluded that the Random Forest method has better analytical accuracy than the Decision Tree method.

According to author Marie-Hélène Roy and Denis Larocque [5], conduct empirical investigations into the resilience of random forests concerning regression tasks. They evaluate the efficacy of six distinct variants of the original random forest methodology, all aimed at bolstering resilience. These adaptations are founded on three core principles: Balance promotes cohesion, Balance reinforces distinction, and Balance instigates positive change. Notably, they opt for the mean (or weighted

average) over the median to derive estimates of individual trees in the initial concept. In the second approach, they utilize the smallest standard deviation from the mean in lieu of least squares for the distribution. Pursuant to the third strategy, they construct the tree based on response points rather than the original outcomes. Through simulated experiments, they compare the performance against synthetic data representing two transmission types alongside 13 real-world datasets. Their findings suggest that all three strategies enhance the robustness of the original random forest algorithm, with cohesive tree grouping often proving more advantageous than separate processes.

According to author Chirwatkar, Rutuja B [6], With the rapid growth in the number of passenger cars and the development of the used car market, used cars are becoming the first choice for customers for various reasons such as financial options, overpriced cars and so on. The online used car platform provides a personal transaction between the seller and the buyer. In this type of systems, the accuracy of used car price evaluation checks whether sellers and buyers can get a more efficient platform. For this paper, he performed a comparative study of regression performance based on supervised machine learning models. Each model is trained using used car market data collected from a German e-commerce website. Car price prediction allows customers to make decisions based on various inputs or factors such as car name, year of manufacture, fuel type, price range, etc.

According to author Stefan Lessmann [7], The article examines statistical models for predicting used car sales prices. An empirical study is conducted to investigate the contributions of different degrees of freedom in the modeling process to forecast accuracy. First, a comparative analysis of alternative forecasting methods provides evidence that random forest regression is particularly effective for forecasting resale prices. It is also shown that the use of linear regression, the predominant method in previous work, should be avoided. Second, the empirical results demonstrate the presence of heterogeneity in resale price forecasting and identify methods that can automatically overcome its detrimental effect on forecast accuracy. Finally, the study confirms that used car dealers have informational advantages over market research agencies that allow them to more accurately predict sales prices. This means that sellers have an incentive to invest in internal forecasting solutions rather than basing their pricing decisions on externally generated residual value estimates.

.

According to author Enis Gegic Becir Isakovic, Dino Kečo, Zerina Mašetić, Jasmin Kevrić [8], investigated car auctions using machine learning techniques (artificial neural networks, support vector machines, and random forests) to predict car prices in Bosnia and Herzegovina. They used data from the Autopijaca.ba portal and compared the performance of the algorithm to determine the most efficient method. Ultimately, they achieved 87.38% accuracy using the test data using the best model in the Java application..

According to author Amit Kumar [9], article explores the evolution of the used car market, driven by comparisons like

car advertisements and influenced by supply and demand. Due to information asymmetry in this market, buyers often lack confidence in purchasing. Kumar aims to solve this problem by using machine learning techniques to estimate the value of used cars. After the data was cleaned and prepared, the main factors affecting the vehicle price were determined through statistical analysis. Four machine learning algorithms were used, with regressor trees performing best, with a test R^2 score of approximately 95%. Additionally, Kumar has developed a weather app to predict car prices, providing users with valuable guidance in determining the value of their cars.

According to author Narayana, Chejarla Venkata and Madhuri, Nukathoti Ooha Gnana and NagaSindhu, Atmakuri and Aksha, Mulupuri and Naveen, Chalavadi [10], highlight the importance of decision-making in business, especially in the context of retail business, especially in second-hand car sales. Their main goal is to develop a prediction model that predicts the selling price of used cars based on values. To achieve this goal, they use machine learning techniques such as random forest regression and feature engineering techniques such as additive tree regression. The results of their approach are very encouraging, showing that the predictive value and consistency of decision trees for the data are successful.

According to Alexander Chernev [11], This study examines consumers' willingness to pay online. Specifically, He compare two inductive strategies: price generation (i.e., “name your price”) and price selection (i.e., “pick your price”). It is generally accepted that consumers prefer price because it provides the most flexibility in expressing their willingness to pay, but this research shows that most consumers prefer to choose price. In our experiment, He show that there is the potential for production costs to vanish due to a lack of accessible information. He also argue that usage cost should be outsourced, and internal usage cost can eliminate the negative effects of production costs and improve consumer choices. These findings support the claim in this study that pre-choice elucidation of benefit costs facilitates consumer choice by creating models consistent with decision work.

According to Listiani, Mariana [12], To be profitable in intense competition, leasing companies need to offer good rental prices. To determine the appropriate price, it is necessary to predict the future prices of used cars By realizing that the car's value is low, you can adjust the rental amount you will pay. A commonly used method for price prediction is multiple linear regression analysis. However, there are many factors that affect the price and complicate this important task. For high-dimensional data, standard regression methods may not be suitable. To solve this problem, support vector regression, a modern data mining technique independent of input dimensions, will be used. The predicted accuracy is then compared to the regression model. In particular, taking ideas from evolutionary research, a complete method for modifying and using SVR was developed. All experiments on machine learning are based on real data from the largest German automobile company.

According to Peerun, Saamiyah, Nushrah Henna Chummun, and Sameerchand Pudaruth [13], The number of vehicles on Mauritius roads increased by 5% last year. In 2014, the National Transportation Administration registered 173,954 vehicles. Therefore, one in six Mauritians owns a car; usually two-hand refurbished and second-hand cars. Artificial Neural Networks Car Price Checking. Thus, data from 200 vehicles from different locations was collected and fed into four different machine learning algorithms. gives slightly better results with network or linear regression. But some estimated values are far from the actual value, especially for more expensive cars. Different network methods and models are still needed to get better predictions.

According to Jian-Da Wu a, Chuang-Chin Hsu a, Hui-Chu Chen [14], This paper presents an expert method for second-hand vehicle price prediction using Adaptive Neuro-Fuzzy Inference System (ANFIS). The system consists of three parts: data collection, cost estimation algorithm and performance evaluation. It is assumed that the most important factors in the current betting market are only the car name, year of production and engine model. In addition, the vehicle's equipment is expected to increase performance at estimated costs. To analyze the results of the proposed ANFIS in cost estimation, the artificial neural network (ANN) with backpropagation (BP) network was compared with the proposed ANFIS due to its variability. ANFIS includes fuzzy logic qualitative approach and adaptive neural network functions. Experimental results show that experts are instructed to use ANFIS more effectively in predicting the second car price.

According to Kishor K, Sharma R, Chhabra M [15], From the information provided, it can be seen that the main focus of this report is education. Our main goal is to educate students. Many important factors are taken into account when developing a model for student prediction. This will help determine the learning area so that the student can be successful in the field of study. Our services use various machine learning algorithms to create predictive models. It is mainly based on linear regression, decision trees, negative Bayes distribution, K-nearest neighbors (KNN) and some improvements in the infrastructure that updates data to make ML easier to understand. The database containing student information is arranged in table format. The row represents the student's name, while each column contains different details about the student, such as family history, gender, medical information, and age. The remaining columns contain the success values that the algorithm is trying to predict. The final report is evaluated by these algorithms and an output is obtained as to whether the student will be successful or not. "Feat Hunch", a student performance tool in ML, aims to connect all students and teachers in a school.

According to Sharma, Jitendra, and Subrata Kumar Mitra[16], This article explores second-hand car prices in India and highlights the average lifestyle and large number of car owners. It uses ordinary least squares and multivariate regression techniques to analyze random values and shows that the complexity of random models increases accuracy.

According to K.Samruddhi, Dr R.Ashok Kumar [17], Estimating the value of used cars is an important and interesting part of analysis. As demand increases in the second-hand car market, the market for buyers and sellers also increases. Reliable and accurate estimates require expertise in this field because the value of the car depends on many important factors. This article presents a machine learning model that uses the KNN (K nearest neighbor) regression algorithm to analyze the value of used vehicles. We train our traffic model using data collected from Kaggle website. Examine the data from this experiment using different training and benchmarks. As a result, the accuracy of the proposed model is approximately 85% and is suitable as a good model.

binary variable which indicates its presence or absence.

• Seller type: same as fuel type in which one-hot encoding is used to convert seller types into binary variables.

• Transmission: By using one hot encoding transmission type can be encode.

• Number of owners: Label Encoding is used for representing categorically. For example (First owner, second owner, etc).

3. METHODOLOGY

A. Data Collection

Data collected for used car price prediction before retrieving Kaggle determined data provided by www.cardekho.com list. This information may include important features such as vehicle make, model, year, mileage, fuel type, transmission and more. Ensuring the quality of the data set includes verifying its reliability, checking for inconsistencies, and confirming that it is representative of the used car industry. After verification, the data set is retrieved in a suitable format (such as CSV or

C. Features Selection

Fig 1.

JSON) and stored for analysis. Metadata information is important for clear, repeatable, detailed information about the work done and the next steps. Additional data augmentation or enrichment techniques can be used to increase the quality and diversity of the data. This data collection process provides a solid foundation for continuous data prioritization, model training and evaluation phase, ultimately leading to the development of accurate and reliable tool applications and price prediction.

B. Data Preprocessing

1. Data Cleaning

During data cleaning, use Pandas' read_csv function to load the dataset. Missing values has been corrected and also resolve the inconsistencies by discarding or removing the results. It handles the duplicate values which involves identifying and removing duplicate entries from the data set to ensure data integrity. This is done using pandas function, which removes rows with same values across all the columns.These duplicates are standardized using string manipulation functions. After removing duplicates it ensures that each data point is unique and prevents bias in modeling tasks.

2. Encoding categorical variables

   • Year of Purchase: Year is represented as a categorical variable. It uses the technique like label encoding in which each year is replaced with a numerical label.

   • Kilometers driven: This can be done similar to the year of purchase using label encoding if its represented categorically.

   • Current Price: label encoding is used. For example: low, medium, high.

   • Fuel type: One- hot encoding is used to covert each fuel type (Petrol, diesel, CNG) into a

   • Correlation coefficients is calculated between numerical features and the target variable.

   • Machine learning model is trained using Random forest regression. X (Feature matrix) and Y (target Variable) is include. X represents the feature matrix, in which it contains the independent variables which is used to predict the target variable (car price). Numerical featuring is consisted and is stored in the data frame. Y represents the target variable, in which it predicts the variable using the features in X. Target variable is price, and is it is stored in the Series data[‘price’]. With X and Y defined, Predictive model can be fitted using this variables.

   • RFE(Recursive Feature Elimination) is employed as the pivotal feature selection technique to discern the most influential features in helping for predicting car prices. First of all dataset undergoes to label encoding for converting categorical variables into numerical labels and facilitates the comptability with RFE algorithm. RFE is applied with the random forest regressor for iteratively rank and features is eliminated based on the importance in predicting car prices. This process continous until the desired number of features is retained. RFE enhances the efficiency of this model, for allowing more accurate and robust prediction of car prices.

Fig 2.

4. Model Selection and Evaluation

Through Feature selection, dataset is splits into the training and testing sets. Following three regression algorithm is used : Linear Regression, Random Forest Regression and Decision tree regression in training the data model. These models are evaluated for testing data by using Mean Absolute Error, Mean Squared Error and R-squared.

1. Linear Regression

Linear Regression is used in the model. It is an statistical method which is used to model the relationship between the dependent variable and one or more than one independent variable. It is commonly used for prediction in various fields which provides insights into the relationships between variables and allowing for predictions of future outcomes. The linear regression model is trained on the given training data which is ‘X’ and ‘Y’ by using fit method. By training the model, the model makes the prediction on both the training set and the testing set through using predict method. Model performance is evaluated for testing data by using score method which computes the coefficient of determination between the predicted car price and actual car price. It finally provide the accuracy which is 73%.

2. Decision Tree Regression

Decision Tree Regression is an algorithm in predicting continuous target variables by splitting the input space into segments. Predictions are made through traversing the tree from root to leaf, target variable is used as prediction in each leaf node. Decision tree regression is also used in this model for comparing the accuracy among the three regression. Also the model is trained on the training data which is ‘X’ and ‘Y’ by using fit method. Decision tree is constructed based on the features and the target variables in training data. The model makes the prediction on both the training set and the testing set through using predict method. Model performance is evaluated for testing data by using score method which computes the coefficient of determination between the predicted car price and actual car price. It finally provide the accuracy which is 88%.

3. Random Forest Regression

Random Forest Regression is an machine learning algorithm which combines multiple decision trees for making predictions for doing regression task. It averages the prediction prediction of all trees in order to obtaining final output. It is a robust and handles complex problems and less prone to overfitting as well. This algorithm is also used in this model in order to comparing them. The n_estimators which is a parameters is set to 15 by specifying the number of trees in the forest. The criterion parameter is also set to Mean Squared Error which indicates the spitting criterion for decision trees. The model is trained on the training data which is ‘X’ and ‘Y’ by using fit method. The model makes

the prediction on both the training set and the testing set through using predict method. It combines prediction of all multiple decision trees to make more robust and accurate prediction for this model. Model performance is evaluated for testing data by using r2 score functions in which it calculates the R-squared (coefficients of determination) between the predicted values and actual values in testing test.nation between the predicted car price and actual car price. It finally provide the accuracy which is 90% which is best accuracy amongs the other two regression models. Hence model is trained in Random forest regression. Here is the output graph of Random Forest Regression Model.

Fig 3.

5. Deployment as a Flask Web App

To deploy the Flask web application using HTML and CSS in Visual Studio Code (VS Code), first set up the Flask environment and ensure that Python and Flask are installed. In the VS Code project directory, Python script is created (e.g. app.py) in which it define the Flask application. In this script, necessary templates and libraries are imported and Flask sample application are created. Paths for different URLs in your application to handle requestsis defined and appropriate HTML structures is generated.

Templates folder in the project directory to store HTML templates for each page of the website is created. These templates can contain dynamic content using Jinja templates. Also static folder with css subfolder is created to hold CSS stylesheets. Section tags to link HTML templates to CSS files are used to implement the templates in web application.

Flask applications is Run locally by running app.py in the terminal. App in VS Code to make sure everything works as expected by debugging and testing Flask application has been deployed to a production or cloud platform is working for public access. Through this Flask web applications is created and deployed using HTML and CSS in VS Code, ensuring a great user experience in web application.

6. Technology used

   1. Flask : It is used for python web frame work for web development. It provides tools, libraries and templates to simplify web development.

   2. Html : Hyper Text Markup Language is used to create web content and defines the content and layout of the user interface.

   3. Css : Cascading Style Sheets is used to create HTML content and controls the layout, colors, fonts and overall appearance of web page.

   4. Python : It is used to create server-side scripts in Flask application. Logic handles data processing and interacting with databases.

   5. Visual Studio Code : In this Flask application is debugged and deployed. It provides syntax highlighting, code completion and continuous improvement program.

   6. Jupyter Notebook : It is an web application which allows to create and share files containing live numbers, equations, visualisation and explanation.

   7. Python Libraries: Many Python libraries are used for data management, machine learning and web development. Some of the commonly used libraries are:

      • pandas: for data management and analysis

      • scikit-learn: for machine learning algorithms and model training

      • matplotlib and seaborn: for data visualization

      • numpy: for used for numerical calculations

      • Request: used for making HTTP requests

      • Flask: used for building web applications

      • Jupyter: used for chat calculation and data analysis

4. CONCLUSION

In summary, Our comparison of a machine learning models for cost prediction demonstrates the superiority of random forest models over linear regression and decision trees. The most accurate prediction found by random forest which indicates its potential use in various automotive industries for example insurance, retail and finance. By leveraging the power of random forest, stakeholders can make more informed decisions about pricing strategies , risk and business analysis. This research contributes to the marketing of vehicle price and highlights the importance ofa advanced machine learning to improve forecast accuracy.

5. RESULT

Our analysis shows significant difference in the prediction accuracy of the three models I.e. Linear Regression, Random Forest Regression and Decision Tree regression. Although linear regression is simple, its ability to capture relationships present in traffic data is limited. Decision trees are slightly more efficient than linear regression, but they tend to overdo it and have a general problem with missing data. In comparison, the random forest model showed better prediction accuracy on various parameters. It shows the ability to integrate the learning process and reduce overfitting. Linear Regression provides 73% accuracy, Decision Tree Regression provides 88% accuracy and Random Forest Regression provides the highest as compared to other two regression which is 90% accuracy. Random

Forest leverages patterns in the dataset, leading to more accurate traffic prediction. Hence model is trained and deployed in Random forest Regression.

Fig 4.

REFERENCES

1. Noor, K., & Jan, S. Vehicle Price Prediction System using Machine Learning Techniques. International Journal of Computer Applications, 167(9), 27-31, 2017.

2. Gongqi, S., Yansong, W., & Qiang, Z. New Model for Residual Value Prediction of the Used Car Based on BP Neural Network and Nonlinear Curve Fit. In Measuring Technology and Mechatronics Automation (ICMTMA), 2011 Third International Conference on (Vol. 2, pp. 682-685). IEEE, Jan 2011.

3. Wu, J. D., Hsu, C. C., & Chen, H. C. An expert system of price forecasting for used cars using adaptive neuro-fuzzy inference. Expert Systems with Applications, 36(4), 7809-781, 2009.

4. P. H. . Putra, A. Azanuddin, B. . Purba, and Y. A. . Dalimunthe, “Random forest and decision tree algorithms for car price prediction”, JUMPA, vol. 3, no. 2, pp. 81–89, Sep. 2023.

5. M. Roy and D. Larocque, “Robustness of random forests for regression” J. Nonparametric Stat., vol. 24, no. 4, pp. 993-1006,July 2012.

6. Chirwatkar, Rutuja B., Price Evaluation of used Car using Machine Learning Technique,Grenze International Journal of Engineering & Technology (GIJET), 2024, Vol 10, p2162, 2395-5287, Jan 2024

7. S. Lessmann and S. Voß, “Car resale price forecasting: The impact of regression method, private information, and heterogeneity on forecast accuracy,” Int. J. Forecasting, vol. 33, no. 4, pp. 864-877, July 2017

8. E. Gegic et al., “Car price prediction using machine learning techniques,” TEM J., vol. 8, no. 1, p. 113, Feb 2019.

9. A. Kumar, “Machine learning based solution for asymmetric information in prediction of used car prices.” International Conference on Intelligent Vision and Computing. Cham: Springer Nature Switzerland, 2022.

10. C. V. Narayana, et al., “Second sale car price prediction using machine learning algorithm” 7th International Conference on Communication and Electronics Systems (ICCES). IEEE, 2022.

11. A. Chernev, “Reverse pricing and online price elicitation strategies in consumer choice,” J. Con. Psychol., vol. 13, no. 1-2, pp. 51-62, 2003.

12. M. Listiani, “Support vector regression analysis for price prediction in a car leasing application,”, 2009.

13. S. Peerun et al., “Predicting the price of second-hand cars using artificial neural networks” in The Second International Conference on Data Mining, Internet Computing, and Big Data (No. August, pp. 17- 21), 2015, Jun..

14. J. D. Wu et al., “An expert system of price forecasting for used cars using adaptive neuro-fuzzy inference,” Expert Syst. Appl., vol. 36, no. 4, pp. 7809-7817, 2009.

15. K. Kishor et al., “Student performance prediction using technology of machine learning” in, Micro-Electronics and Telecommunication Engineering, vol. 373, D. K. Sharma, S. L. Peng, R. Sharma and D. A. Zaitsev, Eds. Singapore: Springer, 2022, 541-551.

16. J. Sharma and S. K. Kumar Mitra, “Developing a used car pricing model applying Multivariate Adaptive regression Splines approach,” Expert Syst. Appl., vol. 236, p. 121277, 2024.4.16.

17. K. Samruddhi and R. A. Kumar, “Used car price prediction using k- nearest neighbor based model,” Int. J. Innov. Res. Appl. Sci. Eng.(IJIRASE), vol. 4, no. 3, pp. 2020-2686, 2020.