Latifa Dwiyanti

Explainable Artificial Intelligence (XAI) has become an increasingly important area of research, particularly as machine learning models are deployed in high-stakes domains. Among various XAI approaches, SHAP (SHapley Additive exPlanations) has gained prominence due to its ability to provide both global and local explanations across different machine learning models. While SHAP effectively visualizes feature importance, it often lacks contextual explanations that are meaningful for end-users, especially those without technical backgrounds. To address this gap, we propose a Python package that extends SHAP by integrating it with a large language model (LLM), specifically OpenAI's GPT, to generate contextualized textual explanations. This integration is guided by user-defined parameters (such as feature aliases, descriptions, and additional background) to tailor the explanation to both the model context and the user perspective. We hypothesize that this enhancement can improve the perceived understandability of SHAP explanations. To evaluate the effectiveness of the proposed package, we applied it in a healthcare-related case study and conducted user evaluations involving real end-users. The results, based on Likert-scale surveys and follow-up interviews, indicate that the generated explanations were perceived as more understandable and contextually appropriate compared to visual-only outputs. While the findings are preliminary, they suggest that combining visualization with contextualized text may support more user-friendly and trustworthy model explanations.

Autosomal dominant polycystic kidney disease (ADPKD) is the predominant hereditary factor leading to end-stage renal disease (ESRD) worldwide, affecting individuals across all races with a prevalence of 1 in 400 to 1 in 1000. The disease presents significant challenges in management, particularly with limited options for slowing cyst progression, as well as the use of tolvaptan being restricted to high-risk patients due to potential liver injury. However, determining high-risk status typically requires magnetic resonance imaging (MRI) to calculate total kidney volume (TKV), a time-consuming process demanding specialized expertise. Motivated by these challenges, this study proposes alternative methods for high-risk categorization that do not rely on TKV data. Utilizing historical patient data, we aim to predict rapid kidney enlargement in ADPKD patients to support clinical decision-making. We applied seven machine learning algorithms—Random Forest, Logistic Regression, Support Vector Machine (SVM), Light Gradient Boosting Machine (LightGBM), Gradient Boosting Tree, XGBoost, and Deep Neural Network (DNN)—to data from the Polycystic Kidney Disease Outcomes Consortium (PKDOC) database. The XGBoost model, combined with the Synthetic Minority Oversampling Technique (SMOTE), yielded the best performance. We also leveraged explainable artificial intelligence (XAI) techniques, specifically Local Interpretable Model-Agnostic Explanations (LIME) and Shapley Additive Explanations (SHAP), to visualize and clarify the model’s predictions. Furthermore, we generated text summaries to enhance interpretability. To evaluate the effectiveness of our approach, we proposed new metrics to assess explainability and conducted a survey with 27 doctors to compare models with and without XAI techniques. The results indicated that incorporating XAI and textual summaries significantly improved expert explainability and increased confidence in the model’s ability to support treatment decisions for ADPKD patients.

In an increasingly modern era, machine learning-based technologies offer solutions to everyday problems. Machine learning enables us to build models that process data and train them to provide predictive results. We call predictions with categorical labels ‘classifications’. We commonly use several basic models for classification, including Decision Trees, Support Vector Machines, Linear Regression, and so on. We can combine these basic models to create an ensemble model, consisting of a boosting algorithm and a bagging model. We will test these two types of algorithms on the profile data of the candidates for the HoR West Java in the 2019 legislative elections. In this paper, we compare the Random Forest bagging algorithm and the Gradient Boosting Decision Tree boosting algorithm. The Decision Tree algorithm is also compared as a baseline. Prior to the comparison, we conduct data pre-processing and data exploration to identify the data features that influence the winning label for feature selection. Furthermore, we address the data imbalance by using Borderline-SMOTE before training the model. We compare the test results based on the F1 score. Our findings reveal that the use of selected features and the Borderline-SMOTE1 sampler significantly improved the F1 score. Additionally, the best-performing model for predicting victory with the given data is the Gradient Boosting Decision Tree, with parameter tuning and using feature selection results, which achieved an F1 score of 0.7445

Promotion abuse fraud is promotion abuse by duplicating accounts to gain an advantage over promotional codes fraudulently. This action inflicts a financial loss on the company. Therefore, this study aims to deal with fraud by developing an application to detect promotion abuse fraud. The application development process includes needs analysis, modeling, and application development. The dataset used for modeling comes from an e-commerce company in Indonesia. The dataset collection stage includes calculating the similarity between accounts using the Levenshtein distance similarity algorithm to get additional features, such as the number of accounts that are similar to an account. At the modeling stage, experiments were carried out using the Random Forest algorithm and a risk scoring algorithm based on machine learning, namely FasterRisk. The FasterRisk has better performance than the Random Forest algorithm shown by a higher F1 score and AUC score. The algorithm also has an advantage in terms of interpretability because it has an output in the form of a more understandable risk score model, so users can understand the factors that are fraud indicators. The FasterRisk prediction result is then deployed into a web application based on user needs.

Advancements in technology have led to the virtual reality development in Indonesia, impacting organizations and businesses positively. Virtual reality offers opportunities for data visualization in the upcoming election, providing insights into candidate wins and vote distribution. This paper develops a virtual reality election data visualization application using Unity and Oculus Quest 2. It includes data acquisition, modelling, rendering, and user interactions. Usability testing was conducted with eight users using task-based scenarios.

In recent years, measuring and monitoring analyte concentrations continuously, frequently, and periodically has been a vital necessity for certain individuals. We developed a cotton-based millifluidic fabric-based electrochemical device (mFED) to monitor glucose continuously and evaluate the effects of mechanical deformation on the device’s electrochemical performance. The mFED was fabricated using stencil printing (thick film method) for patterning the electrodes and wax-patterning to make the reaction zone. The analytical performance of the device was carried out using the chronoamperometry method at a detection potential of −0.2 V. The mFED has a linear working range of 0–20 mM of glucose, with LOD and LOQ of 0.98 mM and 3.26 mM. The 3D mFED shows the potential to be integrated as a wearable sensor that can continuously measure glucose under mechanical deformation.

Indonesia as a democratic country holds general election such as legislative election and presidential election. The General Elections Commissions of Indonesia usually provide useful official data of the election result which grouped by the constituency as well as the province in Indonesia. Moreover, Statistics Indonesia institution has annual statistic data for each province (and in detail up to city/district level per province), such as poverty, prosperity, and education index. Here, we attempt to explore whether there is a correlation between the election result and the province statistic data that can be processed further as the side information for election stakeholders. We focus on (1) providing the platform to perform the big data analytics, (2) identification, selection, and integration of supporting data, (3) visualization of the data on a big data platform; and (4) perform the big data analysis. We use legislative election data (DPR) 2019 in West Java province combined with its demographic data and the party’s ideology to perform the correlation analysis. The result shows that the implementation of big data analysis on such data can offer some side information for election stakeholders.

Today, Information System and Information Technology (IS and IT) become a necessity in almost all sectors of life, including health sector or health services. Health sector need very specific IS requirement, also very high availability of IS systems. Organization need to determine the minimum IS and IT that appropriate to fulfill the need while in other hand should consider best risk, cost, and benefit. There is a mandatory process in Health Care environment to be fulfilled by the IS system. This work try to design a Framework of Information System Architecture for Health Service Organization that built based on Collaborative Care Model (CoCM) and focused on collaboration of component process, human resources, information and technology. The framework consist of IS model, description of component as parts of model (including processes), measurement, and the guideline how to apply the framework. The proposed framework can be used as a reference for any parties that want to build Information System for Health Service Environment.