Munawaroh, Rizki Kurniawati (2023) Implementasi Regresi Logistik dan Random Forest untuk Meningkatkan Efisiensi Penilaian Klaim Asuransi Properti: Studi Kasus Perusahaan Asuransi Umum. Masters thesis, Institut Teknologi Sepuluh Nopember.
![]() |
Text
6032212149-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 September 2025. Download (2MB) | Request a copy |
Abstract
Asuransi properti adalah salah satu bentuk asuransi yang penting karena mampu melindungi properti dari potensi kerusakan. Hal ini membuat peluang pasar asuransi properti menjadi sangat baik, terlebih karena adanya kenaikan penyaluran kredit konstruksi di Indonesia yang berdampak pada kenaikan jumlah permintaan properti. Kenaikan penerimaan premi asuransi properti diiringi dengan kenaikan klaim asuransi properti, sehingga perusahaan asuransi perlu lebih teliti dalam melihat profil risiko tertanggung saat proses penutupan asuransi. Jumlah klaim asuransi properti pada perusahaan asuransi umum (PT XYZ) yang fluktuatif karena proses prediksi klaim yang masih dilakukan secara konvensional sesuai dengan underwriting guideline sehingga menyebabkan subjektivitas analis (underwriting) masih berperan besar dalam proses penutupan asuransi. Oleh karena itu, perusahaan perlu memahami faktor yang mempengaruhi tertanggung melakukan klaim asuransi dengan lebih akurat. Pada penelitian ini, studi yang akan dilakukan adalah peningkatan akurasi klasifikasi klaim pelanggan asuransi (klaim atau tidak klaim) dengan implementasi machine learning, khususnya logistic regression dan random forest. Variabel yang dipilih dalam penelitian dan diduga mempengaruhi keputusan klaim/ tidak klaim adalah kelas konstruksi, okupasi, lokasi objek, zona banjir, zona gempa, penerangan (penggunaan genset/tidak genset), penggunaan sprinkle, fire alarm, hydrant, premi, jenis produk, agen, dan jangka waktu pertanggungan. Penelitian ini melakukan analisis klasifikasi keputusan klaim/ tidak klaim asuransi properti serta mengatasi data yang tidak seimbang dengan penerapan SMOTE (Synthetic Minority Oversampling Technique). Hasil menunjukkan bahwa model klasifikasi random forest sebelum SMOTE memiliki hasil yang paling baik dengan nilai recall atau sensitivity yaitu sebesar 91%. Feature importance dari model klasifikasi yang dibangun diantaranya kelas konstruksi, sumber bisnis, zona banjir, grup okupasi, dan zona gempa. Selanjutnya, perusahaan dapat memprioritaskan penilaian terhadap variabel yang masuk dalam kategori feature importance saat proses penutupan asuransi untuk meminimalisir terjadinya klaim.
===================================================================================================================================
Property insurance is an important form of insurance as it protects property from potential damage. This makes property insurance a very good market opportunity, especially due to the increase in construction lending in Indonesia which has an impact on the increase in the number of property demands. The increase in property insurance premium revenue is accompanied by an increase in property insurance claims, so insurance companies need to be more careful in looking at the risk profile of the insured during the insurance closing process. The number of property insurance claims in general insurance companies (PT XYZ) fluctuates because the claim prediction process is still carried out conventionally by underwriting guidelines, causing the subjectivity of analysts (underwriting) to still play a major role in the insurance closing process. Therefore, companies need to understand the factors influencing the insured to make claims more accurately. In this research, the study will be conducted to improve the accuracy of the classification of insurance customer claims (claim or no claim) with the implementation of machine learning, specifically logistic regression and random forest. The variables selected in the study and thought to influence the decision to claim/not claim are construction class, occupation, object location, flood zone, earthquake zone, lighting (use of generator / no generator), use of sprinkle, fire alarm, hydrant, premium, product type, agent, and coverage period. This research analyzes the classification of property insurance claim / no claim decisions and overcomes unbalanced data by applying SMOTE (Synthetic Minority Oversampling Technique). The results show that the random forest classification model before SMOTE has the best results with a recall or sensitivity value of 91%. Feature importance of the built classification model includes construction class, business source, flood zone, occupation group, and earthquake zone. Furthermore, companies can prioritize the assessment of variables that fall into the feature importance category during the insurance closing process to minimize the occurrence of claims.
Item Type: | Thesis (Masters) |
---|---|
Uncontrolled Keywords: | Klaim Asuransi Properti, Logistic Regression, Random Forest, Variabel Pengaruh; Property Insurance Claims, Logistic Regression, Random Forest, Influence Variables |
Subjects: | Q Science > QA Mathematics > QA76.9.D343 Data mining. Querying (Computer science) T Technology > T Technology (General) > T57.5 Data Processing |
Divisions: | Interdisciplinary School of Management and Technology (SIMT) > 61101-Master of Technology Management (MMT) |
Depositing User: | Rizki Kurniawati Munawaroh |
Date Deposited: | 26 Sep 2023 06:34 |
Last Modified: | 26 Sep 2023 06:34 |
URI: | http://repository.its.ac.id/id/eprint/102876 |
Actions (login required)
![]() |
View Item |