Penerapan metode klasifikasi soft voting berbasis metode boosting dalam deteksi penipuan klaim asuransi mobil

Bastiaan, Matthew Engelbert (2023) Penerapan metode klasifikasi soft voting berbasis metode boosting dalam deteksi penipuan klaim asuransi mobil. Bachelor thesis, Universitas Pelita Harapan.

Preview	Text (Title) Title.pdf Available under License Creative Commons Attribution Non-commercial Share Alike. Download (247kB) | Preview
Preview	Text (Abstract) Abstract.pdf Available under License Creative Commons Attribution Non-commercial Share Alike. Download (878kB) | Preview
Preview	Text (ToC) ToC.pdf Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (580kB) | Preview
Preview	Text (Chapter1) Chapter1.pdf Available under License Creative Commons Attribution Non-commercial Share Alike. Download (577kB) | Preview
	Text (Chapter2) Chapter2.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (721kB)
	Text (Chapter3) Chapter3.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (831kB)
	Text (Chapter4) Chapter4.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (747kB)
	Text (Chapter5) Chapter5.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (563kB)
Preview	Text (Bibliography) Bibliography.pdf Available under License Creative Commons Attribution Non-commercial Share Alike. Download (570kB) | Preview
	Text (Appendices) Appendices.pdf Restricted to Repository staff only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (7MB)

Abstract

Penipuan dalam industri asuransi merupakan masalah serius yang dapat mengakibatkan kerugian finansial bagi perusahaan asuransi dan konsumen. Oleh karena itu, deteksi penipuan asuransi menjadi semakin penting dalam mengatasi masalah ini. Dalam penelitian ini, telah diimplementasikan metode klasifikasi soft voting berbasis boosting untuk memprediksi probabilitas penipuan klaim asuransi mobil. Data yang digunakan dalam penelitian ini diperoleh dari perusahaan asuransi mobil dan berisi informasi tentang klaim asuransi yang diajukan serta status klaim tersebut. Metode soft voting yang digunakan dalam penelitian ini didasarkan pada tiga metode boosting yang telah terbukti efektif, yaitu AdaBoost, GBDT, dan XGBoost. Metode boosting membangun model secara iteratif berdasarkan kesalahan yang diperoleh dari iterasi sebelumnya, sehingga diharapkan dapat mengatasi ketidakseimbangan proporsi kelas pada data. Hasil penelitian menunjukkan bahwa penggunaan metode soft voting dapat meningkatkan performa metode boosting dalam memprediksi probabilitas penipuan klaim asuransi mobil, seperti nilai akurasi hingga 85.3 % dan nilai F1-score hingga 75% . Selain itu, dalam penelitian ini juga dilakukan pemodelan reasuransi yang didasarkan pada hasil prediksi probabilitas penipuan klaim. Hasil penelitian menunjukkan bahwa dengan menggunakan threshold atau ambang batas yang tepat, pemodelan reasuransi dapat membantu meminimalkan kerugian yang ditanggung oleh perusahaan asuransi akibat penipuan klaim asuransi. Dengan menggunakan teknologi ini, perusahaan asuransi mobil dapat mengidentifikasi pola perilaku yang mencurigakan dan mengurangi dampak dari penipuan klaim. Penelitian ini diharapkan dapat memberikan kontribusi yang signifikan dalam mendeteksi klaim-klaim yang diduga sebagai penipuan serta mengurangi kerugian yang ditimbulkan melalui penyusunan produk reasuransi. / Insurance fraud is a serious problem that can cause financial losses for insurance companies and consumers. Therefore, the detection of insurance fraud is becoming increasingly important in this industry. In this research, a soft voting classification method based on boosting techniques has been implemented to predict the probability of fraudulent claims in the automobile insurance domain. The data used in this study was obtained from an automobile insurance company and contains information about the submitted insurance claims and their status. The soft voting method utilized in this research was constructed based on three effective boosting methods, namely AdaBoost, GBDT, and XGBoost. Boosting techniques build models iteratively based on the errors obtained from previous iterations, which is expected to address the class imbalance issue in the data. The results of the study showed that the soft voting method can improve the performance of the boosting techniques in predicting the probability of fraudulent automobile insurance claims, such as the accuracy score up to 85.3 % and the F1-score up to 75 %. Additionally, in this research, reinsurance modeling was also performed based on the predicted probability of fraud. The results demonstrated that by using an appropriate threshold, reinsurance modeling can help minimize the losses incurred by insurance companies due to fraudulent claims. By utilizing this technology, automobile insurance companies can identify suspicious behavior patterns and mitigate the impact of fraudulent claims. The findings of this study are expected to contribute significantly to the detection of suspected fraudulent claims and reduce the associated losses through the development of reinsurance products.

Actions (login required)

View Item