Description

Before making a car insurance claim, the owner must meet several criteria for an insurance claim, not just 1 or 2 people, but many people in 1 day. Therefore, to overcome these problems a classification program was created to identify the owner of the insurance has the right to make a claim or not. The classification program has several stages: data collection, preprocessing, modeling, and evaluation.

Data Collection

At this stage, I collected datasets from Kaggle.com. The dataset is 58592 rows × 44 columns.

Data Preprocessing

At this stage, I did EDA such as statistical description, removing missing values, encoding, data normalization, and correlation.

• Statistical Description

• Removing Missing Values

• Encoding

• Data Normalization

In the data normalization stage, I use MinMaxScaler.

• Correlation

I look for correlations or relationships between features and targets in the correlation process.

Modeling

Classification modeling with machine learning uses Random Forest, Decision Tree, and K-NN algorithms. Dataset division: 30% test data and 70% training data.

As seen in the figure, the modeling results yield 90% for the Random Forest algorithm, 88% for the Decision Tree algorithm, and 93% for the K-NN algorithm.

Evaluation

Conclusion

I pay attention to precision, recall, and f1 score from the modeling and evaluation results. So, I conclude that the Random Forest algorithm best differentiates targets based on the available dataset.