Articles

Explainable AI for Foreign Direct Investment Analysis: Evidence from Central Asia

/in /by

Foreign direct investment (FDI) is an important factor in the economic development of Central Asian countries, where investment flows have traditionally been concentrated in resource-based sectors. In the context of a growing focus on diversification, the need to analyze and study the determinants of FDI is increasing.

This study examines the determinants of FDI inflows in Central Asian countries using machine learning methods (CatBoost) and explainable artificial intelligence (SHAP), and compares the results with a classical econometric approach based on a two-way fixed effects (TWFE) model. Given the limited availability of data, a transfer learning approach is applied: the model is first trained on a group of countries structurally similar to Central Asia and then fine-tuned on the regional sample.

The results show that key macroeconomic factors such as Trade (% of GDP), Current account balance (% of GDP), and several other macroeconomic variables remain significant across both methodologies. At the same time, ML identifies additional regional patterns, such as a higher importance for FDI of determinants including Adjusted savings: carbon dioxide damage (% of GNI), Urban population (% of total population), and Access to electricity (% of population), among others.

The findings indicate that XAI provides interpretable results that are consistent with classical methods and additionally allows for capturing nonlinearities and regional heterogeneity. The study extends the application of ML and XAI in data-constrained Central Asian settings and demonstrates the value of combining econometric and machine learning approaches in the analysis of FDI determinants.

TRAINER – A Content-Based Recommender System of Training Workouts and Nutritional Diets for Fitness Enthusiasts

/in /by

The TRAINER system is a personalized health and fitness solution developed for the Fitness Zone Fitness Center in Antipolo, Philippines, to address the shortcomings of generic, “one-size-fits-all” training programs. By utilizing a content-based recommender system driven by machine learning, the platform integrates individual user profiles—including fitness levels, personal goals, and dietary preferences—with a curated repository of workouts and nutritional plans. The system features a web-based interface for real-time data collection and progress tracking, employing a continuous feedback loop to ensure recommendations remain dynamic and adaptive. Ultimately, TRAINER seeks to enhance client adherence and health outcomes by providing data-driven, context-aware guidance that bridges the gap between limited instructor availability and the diverse needs of fitness enthusiasts.

Sentiment Analysis Based on Questionnaires: A Case Study on the Use of Induction Stove

/in /by

Indonesia’s reliance on subsidized Liquefied Petroleum Gas (LPG) for household cooking places a significant burden on the national energy subsidy budget and increases dependence on imported fossil fuels. As part of the clean energy transition strategy, the Indonesian government has promoted the conversion from LPG stoves to electric induction stoves. However, public acceptance and actual post-use experiences at the household level remain diverse and insufficiently examined empirically. This study aims to analyze public sentiment toward induction stove use based on post-adoption user reviews to identify factors that encourage interest and reveal existing adoption barriers.

This study employs a machine learning–based sentiment analysis approach using primary data collected through open-ended questionnaires distributed to induction stove users. A total of 265 valid textual responses were analyzed. Text preprocessing was conducted using Python with the NLTK and Sastrawi libraries, including data cleaning, case folding, tokenization, stopword removal, stemming, and duplicate removal. Sentiment classification was performed using the Term Frequency–Inverse Document Frequency (TF-IDF) method and the Naive Bayes algorithm, while WordCloud visualization was applied to identify dominant keywords.

The results indicate a relatively balanced sentiment distribution, with positive sentiment accounting for 33.6%, neutral sentiment 32.5%, and negative sentiment 34.0%. Positive sentiment is mainly associated with energy efficiency, safety, and ease of use, whereas negative sentiment is driven by concerns regarding initial costs and electricity dependence. Neutral sentiment reflects an evaluative phase among users. These findings provide empirical insights to support user-oriented policies and strategies for accelerating the sustainable adoption of induction stove technology in Indonesia’s clean energy transition.

Behavior-Based Purchase Intent Prediction in E-Commerce: A Machine Learning Approach

/in /by

This study investigates the use of machine learning to predict user purchase intentions based on behavioral data in a multi-category e-commerce platform. By analyzing seven months of user interaction logs—comprising product views, cart additions, and purchases—the research applies feature engineering to generate variables such as event weekday, product category levels, session activity count, and cart-to-view ratios. Four classification models were developed and evaluated: logistic regression, decision tree, random forest, and gradient boosting. Among these, the Random Forest algorithm outperformed the others, achieving the highest accuracy and F1-score, effectively balancing precision and recall. The results demonstrate that machine learning can reliably predict purchase intent and support more targeted marketing, personalized recommendations, and improved conversion strategies in e-commerce environments.

A Review of AI-powered Diagnosis of Rare Diseases

/in /by

The diagnosis of rare diseases presents significant challenges due to their low prevalence, complex symptomatology, and the scarcity of specialized knowledge. However, advancements in Artificial Intelligence (AI) offer promising solutions to these challenges. This review explores the current state of AI-powered diagnostic tools for rare diseases, focusing on the methodologies, algorithms, and platforms utilized in this emerging field. We examine how AI technologies, such as machine learning, deep learning, and natural language processing, are being integrated into clinical practice to enhance diagnostic accuracy and speed. The research also provides the examples that highlight the successes and limitations of AI in this domain, providing insights into how AI can be harnessed to improve patient outcomes in rare disease diagnosis and management.

Gesture Plus: A Novel Approach to Enhance Interactive Media

/in /by

 GesturePlus is a comprehensive human- computer interaction system that integrates hand gestures, voice commands, and a chatbot for seamless handling. Through the use of computer vision and machine learning, GesturePlus recognizes hand gestures through image pre- processing, fingertip detection, and real-time classification with high accuracy. GesturePlus finds great use within sterile environments or for people with limited mobility in an intuitive manner that replaces the traditional input devices.

Artificial Intelligence and Machine Learning- Driven Pharmaceutical Industry

/in /by

Artificial Intelligence (AI) and Machine Learning (ML) are revolutionizing the pharmaceutical sector at every stage—drug discovery, development, regulatory affairs, quality control, and post-marketing surveillance. These technologies improve data processing, accuracy, and timelines by using complex algorithms and large volumes of healthcare data. AI helps in drug target identification, drug design, prediction of toxicity, and pharmacokinetics modeling, as well as improving regulatory processes and pharmacovigilance. Though they have their benefits, there are still challenges such as data privacy, algorithmic bias, explainability, and accountability. Regulatory structures and ethical implications need to keep pace so that AI can be used safely and fairly in pharmaceuticals. This article discusses the existing applications, advantages, risks, and future possibilities of AI and ML in transforming drug development and healthcare outcomes.

Optimization of Wireless Mesh Networks for Disaster Response Communication

/in /by

Wireless Mesh Networks (WMNs) have emerged as a resilient and adaptable solution for disaster response communication, offering self-healing and self-organizing capabilities that ensure uninterrupted connectivity in emergency scenarios. Traditional communication infrastructures often fail due to network congestion, power outages, and physical damage during disasters, necessitating an optimized approach for rapid and reliable data transmission. This study presents an AI-optimized WMN framework aimed at enhancing network performance by improving packet delivery ratio (PDR), reducing end-to-end delay, optimizing energy consumption, increasing network throughput, and strengthening security. Simulations conducted in MATLAB Simulink compare the performance of AI-optimized routing with conventional protocols such as AODV (Ad hoc On-Demand Distance Vector) and OLSR (Optimized Link State Routing). Results demonstrate that AI-optimized routing achieves a 15.5% higher PDR, 43% lower delay, 49% increased throughput, and 30% reduced energy consumption compared to traditional approaches. Furthermore, an AI-driven Intrusion Detection System (IDS) improves network security by increasing attack detection accuracy to 94.6% while reducing false positive rates to 5.2%. The findings highlight the significance of AI-based routing optimization in disaster scenarios, ensuring robust, energy-efficient, and secure communication for first responders and affected communities. Future research will explore hybrid AI-blockchain security mechanisms, 5G and satellite network integration, and real-world experimental validation to further enhance WMN resilience in extreme disaster conditions.

Characteristics of Rainfall Influence Indian Ocean Dipole (IOD) and Madden Julian Oscillation (MJO) Phenomena Based on Machine Learning in Deli Serdang Region

/in /by

Rainfall in the Deli Serdang region is influenced by global climate phenomena. This study aims to determine the characteristics of rainfall based on machine learning due to the simultaneous occurrence of IOD and MJO in the Deli Serdang region. This study uses a descriptive method and Pearson correlation analysis using rainfall, IOD, and MJO data. The results of the study with machine learning showed that the accuracy value of the SVM model was 56.16% and when the MJO was strong and the IOD was positive in January – December 2024 in the Tuntungan region, the highest was 258 mm and the lowest was Bandar Khalipa 167 mm. Strong MJO and Negative IOD were found in December 2022, the highest area was Sibiru-biru 264 mm and the lowest was 146.16 mm. Weak MJO and Positive IOD in the low-lying Bandar Khalipa region were 140 mm. Dry months can be predicted using several indicators, including the MJO (Madden-Julian Oscillation) and IOD (Indian Ocean Dipole). However, dry months are more often predicted using the IOD indicator. IOD has a significant influence on rainfall in Indonesia, especially in eastern Indonesia. When IOD is in a positive phase, rainfall in Indonesia tends to decrease, increasing the possibility of a dry month. MJO has a greater influence on rainfall on a shorter time scale, such as weekly or monthly. MJO can affect rainfall in Indonesia, but its influence is not as great as IOD in predicting dry months.

An Explainable Artificial Intelligence (XAI) Methodology for Heart Disease Classification

/in /by

Heart disease continues to be one of the predominant contributors to morbidity and mortality on a global scale, underscoring the imperative for early and precise diagnosis to enhance patient outcomes. Machine Learning (ML) has emerged as a formidable instrument in the classification of cardiovascular diseases, utilizing intricate clinical datasets to discern patterns that conventional statistical methodologies may fail to detect. Nevertheless, notwithstanding their robust predictive capabilities, numerous machine learning models function as black-box systems, exhibiting a deficiency in transparency regarding their decision-making processes. The absence of interpretability presents a considerable challenge in clinical environments, where trust, accountability, and elucidation are of utmost importance for medical professionals. In order to tackle this issue, we propose a methodology for heart disease classification that is grounded in Explainable Artificial Intelligence (XAI). This approach incorporates interpretable machine learning models to improve diagnostic transparency and reliability. Our framework conducts an evaluation of various classifiers, including Support Vector Machine (SVM), Gradient Boosting (GB), Extreme Gradient Boosting (XGBoost), Multi-Layer Perceptron (MLP), and LightGBM. This assessment is based on essential performance metrics, namely accuracy, precision, recall, F1-score, and AUC-ROC. Furthermore, SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-Agnostic Explanations) have been integrated to enhance the interpretability of the model. The experimental findings indicate that XGBoost surpasses alternative models, attaining the highest classification accuracy of 92% and an AUC-ROC score of 0.93, all while preserving interpretability. This study underscores the significance of incorporating Explainable Artificial Intelligence (XAI) techniques within medical AI applications. It advocates for the adoption of transparent, interpretable, and clinically dependable machine learning methodologies to enhance clinical decision-making and optimize patient outcomes.