ItemBook Chapter

Prediction of health insurance premium using bidirectional long short-term memory network with local interpretable model-agnostic explanations

Title
Prediction of health insurance premium using bidirectional long short-term memory network with local interpretable model-agnostic explanations
Creator
Jangid, Harsh; Kannan, M.; Dua, Arush; Jaiswal, Kush; Singh, Gurinder; Agrawal, Rayan
Description
This research proposes an application of deep learning techniques towards the prediction of insurance premiums using ConvLSTM, BI-LSTM, and CNN-LSTM models. Nowadays, Insurance is becoming more sophisticated, there is a need for better models that predict premiums so that risk factors that can be properly valued. The aim of this study is to improve the accuracy and reliability of insurance premium prediction using deep learning methods. The main challenge is the shallow traditional models, whose capturing of temporal dependencies is ineffective and results are not explainable resulting in very few stakeholders having any trust to the predictions. To solve this, this study compared three models: ConvLSTM model, BI-LSTM and CNN LSTM. Of these, the BI-LSTM model was the most effective because it was able to learn bidirectional sequential patterns. These patterns were enhanced using L2 regularization, dropout and dense layers to improve generalization. The dataset used comes from a Kaggle repository, which contained actual insurance data incorporating age, BMI, region and smoking as attributes. Results showed that BI-LSTM had performed the best as compare to other models in terms of accuracy and loss minimization. Important findings highlighted features such as age, smoking, and BMI as pivotal to estimating premiums. Also, to make the model explainable, we incorporated Explainable AI using LIME which delivers interpretable explanations by showing and visualizing the most important features for single predictions. 2026 selection and editorial matter, K. V. Sambasivarao, and Anasuya Sesha Roopa Devi Bhima; individual chapters, the contributors. All rights reserved.
Source
Artificial Intelligence, Computational Intelligence, and Inclusive Technologies: Proceedings of International Conference on Artificial Intelligence, Computational Intelligence, and Inclusive Technologies (ICRAIC2IT 2025);pp.621-630
Date
01-01-2026
Publisher
Taylor and Francis
Subject
Deep learning; Health insurance; LSTM; Machine learning; XAI
Coverage
Jangid H., Department of Computer Science, CHRIST University, Bengaluru, India; Kannan M., Department of Computer Science, CHRIST University, Bengaluru, India; Dua A., Department of Computer Science, CHRIST University, Bengaluru, India; Jaiswal K., Department of Computer Science, CHRIST University, Bengaluru, India; Singh G., Department of Computer Science, CHRIST University, Bengaluru, India; Agrawal R., Department of Computer Science, CHRIST University, Bengaluru, India
Rights
Restricted Access; Hardcopy may be available in the library
Relation
ISBN: 978-104089003-5; 978-104124091-4;
Format
online
Language
English
Type
Book chapter

Citation

Jangid, Harsh; Kannan, M.; Dua, Arush; Jaiswal, Kush; Singh, Gurinder; Agrawal, Rayan, “Prediction of health insurance premium using bidirectional long short-term memory network with local interpretable model-agnostic explanations,” CHRIST (Deemed To Be University) Institutional Repository, accessed June 17, 2026, https://archives.christuniversity.in/items/show/24469.

Output Formats