Description:Field of Invention
The present invention is a desktop application which helps user to predict the price of a used car based on the features of the used car such as Brand, model, odometer readings, Transmission Type, Fuel Type, etc.
The Objectives of this Invention
The main objective of this invention is to predict the price of the used car. Initially data will be collected from open-source website which is cars24, after that data cleaning is performed to remover unwanted values and next with the help of python sklit-learn models the model will be trained.
Background of the Invention
There are some websites such as cars24.com, carsdekho.com which helps user to predict the price of used car but in-order to predict the price the user should give their phone number. The information will go the staff where they will try to contact us and try to help us in sell our car. For a user who just wants to know the price but getting a call and replying them is a hectic process. Also, the algorithms used in websites may not be accurate to predict the price.
Upon study on this paper (Car Price Prediction Using Machine Learning, IJCSE, Vol.-7, Issue-5, May 2019 ), it has been found that the Machine learning Algorithm which has been used for implementation of model is K-Means Algorithm and Decision Tree Algorithm. The features that have been used to predict the price of a car are Kilometers Traveled, Fiscal Power, Year of registration, Fuel Type. There is no information about the Datasets which has been used for training. Hence, we can assume that the datasets have been taken from the online source. In this article (Car Price Prediction using Machine Learning Techniques, TEM Journal. Volume 8, Issue 1, February 2019), it has been found that the Machine Learning Algorithm which has been used for implementation of model is Random Forest Algorithm. The features that have been used to predict the price of a car are Brand, Model, Fuel Type, HorsePower, Year of Manufacture, Miles, Leather, Cruise Control. There is no information about the Datasets which has been used for training. Hence, we can assume that the datasets have been taken from the online source.

In (JP2008/4241882B1), the current invention is a system for predicting the residual value or residual value of a used object or vehicle with accuracy. An article resale value forecasting equipment comprises an article resale value calculating processor, a plurality of item names, a used article pricing for each item type, a news item price for every piece type, and then a year whereby the used article cost is implemented. Another invention (US2017/10885562B2), an extracting device set to receive datasets from application-specific file source libraries can be included in a visual exploration tool for auto manufacturing with networking cryptography, data purification, and predictions. A car alert repository can also be included in the tool, which obtains vehicle information from multiple extraction sources.

Summary of the Invention
We want to achieve a better result with fewer errors and maximum precision. The focus of the strategy is to forecast the price of a used car. Data will be acquired from an open-source website called cars24, after which it will be cleaned to eliminate any undesired values, and the model will be trained using Python sklit-learn models. The best models will be included into the User Interface, with the result calculated as the average of the models' forecasts. The allows the user to interact with that as well here and get the cost of the used car based on the information provided by the client.

Detailed Description of the Invention
The proposed invention is made up of model which uses ensemble machine learning models which are related to decision trees. The Ensemble machine learning models that are been used for training of model are Random Forest Regression, XGBoost Regression, AdaBoost Regression. The Dataset will be collected by the model itself i.e from the car selling website such as Cars24 [https://cars24.com] which records over 5000+ dataset values. To help users interact with the model a UI will be developed. The UI will be built by PyQT5 tool. The UI will take inputs i.e features/details of a car from the user and then displays the predicted price. The algorithms that have been used are easy to Grasp and plot on graph or visualize and describing the Machine learning model. Model will be trained on live data so that it can predict the used car price on today’s date. The UI model is built on kivy so the application can run on Desktop of any os which can be windows, linux or mac i.e it is portable.

The architecture is divided into two parts; Implementation of Model; User Interface. The implementation is divided into two phases: Implementation of model, User Interface. In Implementation of model phase, we have Data collection and pre-processing of the data. The model will get trained and tested. In User Interface phase the user will interact with model by loading the features of the car and gets the price. The Detail column has been split into 3 columns which are EngineType, Milagecovered, InHand as well as the Product column is splitted into BrandName and Model. In order to deal with null values, the complete row will be removed. After the training has done the model will be tested with Testing Data and their R2 Scores will be compared among each other. The Best R2 Score model will be selected among the 3 models and to save that model it will be stored in a pickle file to be used in further deployment of User Interface. The UI will interact with user and predicts the price. According to the graph we can see all the models are giving 85+ accuracy so the result of prediction will be the average of the 3 model results.

3 Claims & 1 Figure
Brief description of Drawing
In the figure which is illustrate exemplary embodiments of the invention.
Figure 1, the Process of NextCar Predictor , Claims:The scope of the invention is defined by the following claims:

Claim:
1. A system for predicting the price of used cars based on current market value, said system comprising the steps of:
a) The system has the training models (1) with three different algorithms, then pickle file (2) will be there for selecting the best models.
b) Feeding those models to the UI so that the user can interact and give the inputs. The UI will again predict (3) the price according to those inputs.
2. As mentioned in claim 1, the users can get price of the used car based on current market value. Then from the datasets, the values are trained by using three different ML algorithms.
3. According to claim 1, the users can install this application and predict the car prices easily.