onedayxzn / Prediksi-Harga-Mobil-Volkswagen

Domain Proyek.

Competition in the industrial sector is getting tighter and more competitive every year, so companies are indirectly required to continue to be creative and innovate in offering products or services to be sold. so that every company, whether offering products or services, has the aim of staying alive and growing. Or the consumers that the company already has do not move to other products and to attract new consumers to consume products that have been produced by a company. consume products that have been produced by a company. Along with this idea, the marketing concept has also developed. Marketing activities are now focused on satisfying consumer needs.

Marketing itself must be thought out long beforehand, to be more precise on target consumers. Because potential consumers will consider various factors, including value factors, brand image and trust in the brand before choosing a product that can provide the highest satisfaction to consumers. This is also experienced in the automotive industry business sector, the emergence of new competitors with marketed products causes more consumer choices to choose and buy which products will be consumed to meet their needs Automotive itself is an industry engaged in transportation that produces community needs in the form of vehicles as a means of transportation. One of them is a four-wheeled vehicle, namely a car, which has become a necessity that cannot be separated from human life. There are several things that encourage high automotive sales in Indonesia, namely the increasing number of upper middle class in Indonesia. The increase in per capita income encourages many people to experience an increase in welfare and enter the category of groups with lifestyles that are starting to be exclusive, and Indonesia has a lot of human resources (HR), thus allowing a wide opportunity for the domestic manufacturing industry to develop further in supporting the domestic automotive industry.

In general, people buy cars as a means of mobility from one place to another, and as a means of transporting goods in their daily lives. In the midst of the development of the car industry today, which is increasingly competitive, competition between car manufacturers is getting tighter, and makes car manufacturers to always develop their products and try to capture existing market share.

• The problem that must be solved is how the car product can compete in the market, and can predict the price of the product.

Business Understanding.

Problem Statements

1. What features have an impact on car prices?
2. What algorithms are suitable for price prediction?

Goals

Make car price predictions so that buyers can easily calculate the price of cars in the future

Solution statements

Solution Statements that will be done is by applying 3 Machine Learning algorithms, which are:

• Random Forest.
  Random forest algorithm is one of the supervised learning algorithms. It can be used to solve classification and regression problems. Random forest is also a frequently used algorithm because it is quite simple but has great stability.
• Boosting Algorithm.
  The boosting algorithm works by building a model from the training data. It then builds a second model that corrects the errors of the first model. Models are added until the training data is well predicted or has reached the maximum number of models to add.
• K-Nearest Neighbor.
  KNN is a relatively simple algorithm compared to other algorithms. The KNN algorithm uses 'feature similarity' to predict the value of each new data point. In other words, each new data point is assigned a value based on how similar it is in the training set.

By applying the 3 algorithms above, it will be sought which algorithm has the smallest error rate or error. So that the predicted price is close to the original price.

Data Understanding

The data or datasets used in this machine learning project are the following 100,000 UK Used Car Data set that was obtained from the kaggle site. The dataset link can be seen from the following link vw.csv

The variables in the House Sales in King County, USA dataset are as follows:

• model : is a list of models that exist in Volkswagen cars.
• year : is a list of the year the car model was released.
• Price : lists the price of the car (in euros).
• transmission : is a list of transmissions on the car.
• mileage : is a list of mileage that can be traveled.
• fuelType : lists the type of fuel used.
• tax: is a list of taxes
• mpg: is a list of fuel efficiency.
• engineSize: is a list of engine capacities.

Data Preparation

The data preparation used by me is:

• Data selection: selecting data whether the data is empty or not, if there is empty data then I will delete it.
• Dividing data into training and test data: to divide data for training and test.
• Handling Outliers = handling samples whose values are very far away.
• Using OneHotEncoder = encoding the model, transmission, and fueltype features.
• Dimensionality Reduction with PCA = reduces the number of features while maintaining the information in the data.
• Standardization = Helps make data features into a form that is easier for algorithms to process.

Modeling

The modeling process I did on this data was to combine three machine learning algorithms and then find the best performance of the three machine learning algorithms.

This is the result of the best model from the three algorithms that I used

It can be seen from the bar chart above from the three algorithm models that I used, that Random Forest has the smallest error value.

This is the result of the model I used

From the table above, it can be seen that prediction using Random Forest gives the closest results compared to the other two models.

Evaluation

The evaluation metric used to measure the performance of the model is the mse (Mean Squared Error) metric, as my chosen case is a regression case.

MSE basically measures the mean squared error of our predictions. For each point, it calculates the squared difference between the prediction and the target and then averages those values.

The higher this value, the worse the model. The MSE value is never negative, as we square the individual prediction errors before summing them, but would be zero for a perfect model.

Benefit: Useful if we have unexpected values that we have to care about. A very high or low value that we have to pay attention to.

Weaknesses: If we make one very bad prediction, squaring will make the error worse and that might make the metric tend to overestimate the badness of the model. That is a very problematic behavior if we have noisy data (i.e., data that for whatever reason is not completely reliable) even a "perfect" model may have a high MSE in that situation, so it becomes difficult to judge how well the model is performing. On the other hand, if all errors are small, or more precisely, smaller than 1, than the opposite effect is felt: we can underestimate the badness of the model.

The formula of MSE metric is as follows

Description:
N : Number of datasets.
yi = actual value.
y_pred = predicted value.