Journal of Advanced Research in Embedded System

A Comparative Study of ARIMA and LSTM Models for Short-Term Forecasting of the El Nino Modoki Index

2025-10-20T12:15:49+00:00

To get a handle on global climate patterns, we need to accurately forecast the El Niño Modoki Index (EMI). This study compares two methods for predicting the EMI's short-term behaviour based solely on its historical data. We pitted a classic statistical method, the AutoRegressive Integrated Moving Average (ARIMA) model, against a modern deep learning approach using a Long Short-Term Memory (LSTM) network. We trained both models on monthly EMI data from 1982 to 2022 and then tested how well they could predict the period from 2023 to 2025.
Our results showed that the ARIMA(2,0,0) model works as a solid, understandable baseline, capturing the main movements of the index with a Root Mean Squared Error (RMSE) of 0.4338 and an R-squared (R²) of only 0.0533. The LSTM network, however, was much better at handling the quirky, non-linear nature of the data, leading to a far more accurate forecast with an RMSE of just 0.0820 and an R² of 0.9658. Ultimately, while a simple ARIMA model is useful as a benchmark, our work makes it clear that LSTM networks can offer a major leap forward in forecasting accuracy for complex climate indicators like the EMI.

Comparison of Machine learning model with Explainable AI: Applicable to Dementia

2025-10-20T13:05:39+00:00

In the healthcare domain, clinical practice needs effective models with flourishing interpretability to address any issues, cases like Dementia, in which diagnosis needs a proper explanation for such urgent problems, need an accurate model with effective interpretability. In medical practice implementation of Machine Learning (ML) models presents difficulties because of a lack of clarity on how particular results are derived, even though their
outcomes are accurate. Traditional ML models with Explainable Artificial Intelligence (XAI) and without XAI by using Open Access Series of Imaging Studies (OASIS) dementia datasets are used to find out which has more interpretability to show the comparative analysis. SHAP (SHapley Additive exPlanations)/Local Interpretable Model-Agnostic Explanations (LIME) of XAI are used to provide explanations, whereas Metrics like Accuracy, Recall, Precision, F1-score, and AUC are used to evaluate the base model whose results are then compared with
other metrics to find the importance of interpretability in the model to overcome the gap between ML model and its implementation in clinical practice. The Traditional ML model provides good anticipating accuracy with an Area Under Curve (AUC) up to 0.94, but incorporating ML with the XAI model together gives better clinical results and enables medical professionals to build trust in predictions made by models. This clarifies the decision-
making capabilities of ML models, eliminating risk factors. Thus, this study describes the need for an effective way to diagnose diseases is not only through good models with high accuracy, but also with models providing interpretability and clarity on prediction. Therefore, leading to an analysis that indeed helps in addressing gaps in implementing models applicable to the medical domain.

Measuring ENSO Predictability Sources Using Machine Learning Techniques

2026-02-02T15:07:55+00:00

Important global climate anomalies, such as the El Niño-Southern Oscillation (ENSO), can impact socio-economic systems, water resources, and agriculture. Strategies for disaster preparedness and climate adaptation rely on accurate ENSO event predictions. Here, we use machine learning techniques to investigate potential causes of ENSO prediction. The ENSO indices (Niño 3.4 and Niño 4) were modelled using historical records of sea surface temperature (SST), sea level pressure (SLP), and subsurface ocean temperatures. To unravel the data’s patterns, including its non-linear linkages and temporal dependencies, we employed Random Forest (RF), Gradient Boosting (GB), and Long Short-Term Memory (LSTM) networks. To evaluate model performance, we used RMSE, MAE, and correlation coefficients for performance metrics, as well as feature importance metrics and seasonal analysis to evaluate phase-dependent predictability (i.e., modelling winter ENSO and summer ENSO). The results indicate that the LSTM model yields performance levels superior to the tree-based models and predicts the highest levels of ENSO prediction accuracy; the strongest predictors were SST anomalies in the Niño 3.4 region and subsurface temperatures. In terms of seasonal predictability, we found that ENSO events during winter months were more predictable than summer months, which aligns with phase-locking behaviour. Overall, this research shows that machine learning can provide reliable understanding of ENSO dynamics and identify the important climate drivers, which together provides a step towards better forecasting and early warning capacity.

AI-Powered Disease Diagnostic Predictive Model using Neural Networks

2026-02-03T12:04:32+00:00

The digital transformation of healthcare has seen an increasing reliance on Artificial Intelligence (AI) and Machine Learning (ML) technologies to support diagnostic and decision-making processes. In a world where access to healthcare services remains uneven—especially in remote or economically underdeveloped regions—technology is being leveraged to fill the accessibility gap. The rise of intelligent applications has not only enhanced the accuracy of diagnostics but has also enabled faster, scalable, and cost-effective solutions. This research introduces a lightweight neural network web application that identifies patterns within symptoms to suggest a probable disease. With a main focus on accessibility, affordability, and adaptability, the core of this system is a deep learning model trained on a dataset consisting of 5,000 symptom-disease mappings covering 41 unique diseases. The model with regularisation has achieved an outstanding 96.54% accuracy. The neural network with users, whether healthcare professionals or general individuals, can interact with the application to input symptoms and receive a disease prediction within seconds. This serves as an initial assessment tool, prompting users to seek professional advice if necessary. The system is designed to be lightweight using TensorFlow Lite, making it deployable even on low-end devices. It is hosted online to ensure ease of access and is free of cost, promoting inclusivity. The incorporation of a feedback mechanism—where users can correct wrong predictions—adds another layer of intelligence by laying the groundwork for reinforcement-based learning in future versions.

AI-Driven Integrated System for Climate Disaster Mitigation: An Early-Warning Architecture for India

2025-10-27T04:37:57+00:00

India faces recurrent climate hazard floods, cy- clones, wildfires, heatwaves, and droughts whose intensity and frequency have accelerated under anthropogenic climate change. Although conventional Early Warning Systems (EWS) have significantly improved in recent decades, they continue to face limitations in temporal resolution, spatial accuracy, and impact- based forecasting. This paper presents a comprehensive AI- driven multi-hazard EWS architecture for India, integrating meteorological, hydrological, and socio-economic data using Ma- chine Learning (ML) and Deep Learning (DL) techniques. By fusing multi-source satellite imagery, gridded observations, and a curated Vulnerability Database, the system aims to improve predictive precision and deliver hyperlocal alerts. The frame- work emphasizes hybrid architectures Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) networks, and Transformers to achieve scalable, real-time, and explain- able hazard predictions. The proposed design aligns with India National Disaster Management Plan (NDMP-2023) and the UN Sustainable Development Goal 13 on Climate Action.

AI-Driven Farm-Guard Protection System from Offensive Entry of Nilgai

2025-10-27T04:27:45+00:00

Agriculture is the backbone of any country's growth, and it is the utmost duty of every people in that country to save it and help in cultivating, but farmers have no such technology to save the crops from being destroyed by animals. Animal intrusion is a major threat to the productivity of the crops; it becomes a major concern for farmers and a threat to the crop fields. To protect their crops from animals like Wild Boars, Goats, and Monkeys, Nilgai (Boselaphus TragoCamelus) etc. farmers lead to take preventive measures such as installing fences which becomes a danger for animals too. The proposed system has been developed to create a defensive framework for the crop fields without making harm to the animals using Deep Learning model and IoT sensors. Upon detecting a wild animal system triggers multiple deterrent mechanisms, including the activation of a buzzer, water sprinkler and flashing of light. The proposed system helps the farmer to protect their crops from damage due to wild animals without any harm to them which also helps in preserving biodiversity.

Spatiotemporal Prediction of Cloudburst Vulnerability Zones in Uttarakhand Using ERA5 Reanalysis (1960–2024)

2026-01-17T07:40:51+00:00

A Review of Post-Harvest Ripening Techniques and Their Correlation with Fruit Quality

2026-01-17T08:54:33+00:00

This paper reviews different fruit ripening methods and how they affect fruit quality and consumer health. Calcium carbide is a commonly used ripening agent in developing coun- tries. It speeds up ripening but leads to nutrient loss, shorter shelf life, and higher levels of heavy metals and toxic residues. Natural ripening maintains moisture, organic acids, minerals, and sensory quality but takes more time and does not meet the requirements of the fruit supply chain. There are certain biological agents such as African bush mango fruits and Jatropha curcas leaves that provide safer options by maintaining higher nutrient levels and less contamination. Traditional smoke-based methods cause fast ripening but add polycyclic aromatic hydrocarbons and lead to uneven ripening. Ethylene treatment is the only globally accepted safe artificial ripening method as it achieves even ripening with significant nutritional retention. Ethephon assisted with vacuum speed up ripening but might lower vitamins and carotenoids. Other agents like potash, ethylene glycol, lauryl alcohol, propylene, and methyl jasmonate vary in safety and effectiveness. Overall, artificial methods speed up ripening but often reduce nutritional value and safety which highlights the need for regulating the process of artificial ripening.

Crop Yield Prediction: A Comprehensive Review of Machine Learning and Deep Learning Approaches

2026-01-19T10:10:44+00:00

Predicting crop yields accurately is essential for farm management, policymaking, and food security. With an emphasis on applications in India, this paper examines current developments in machine learning (ML) and deep learning (DL) techniques for crop yield estimation worldwide. The current study examines different obstacles (data scarcity, model transferability, interpretability), input variables (weather, soil, satellite indices), model types (regression, tree ensembles, neural networks, hybrid architectures), and future possibilities (transformers, multimodal fusion, IoT). In addition, this paper points out gaps and suggests recommended practices for further study.

AI-Driven Farm-Guard Protection System from Offensive Entry of Nilgai

2026-01-19T11:04:09+00:00

Agriculture is the backbone of any country's growth, and it is the utmost duty of every citizen in that country to save it and help in cultivating, but farmers have no such technology to save the crops from being destroyed by animals. Animal intrusion is a major threat to the productivity of the crops; it becomes a major concern for farmers and a threat to the crop fields. To protect their crops from animals like wild boars, goats, monkeys, nilgai (Boselaphus tragocamelus), etc., farmers are led to take preventive measures such as installing fences, which becomes a danger for animals too. The proposed system has been developed to create a defensive framework for the crop fields without causing harm to the animals using a deep learning model and IoT sensors. Upon detecting a wild animal, the system triggers multiple deterrent mechanisms, including the activation of a buzzer, water sprinkler and flashing of light. The proposed system helps the farmer to protect their crops from damage due to wild animals without any harm to them, which also helps in preserving biodiversity.

IoT-Driven Disaster Management Systems: A Framework for Early Warning, Mitigation, and Resilient Recovery

2026-02-02T09:43:24+00:00

The Internet of Things (IoT) has emerged as a transformative paradigm reshaping disaster management through intelligent connectivity, real-time sensing, and automated decision-making. This paper explores the evolution of IoT from early networked devices to its present role in enabling resilient, data-driven disaster response frameworks. It analyses traditional early warning mechanisms and contrasts them with IoT-enabled systems that integrate sensors, cloud platforms, and AI analytics for proactive mitigation and rapid recovery. Applications across domains, including floods, landslides, wildfires, earthquakes, and counter-terrorism, demonstrate how IoT enhances situational awareness, resource coordination, and risk reduction. Real-life case studies from Jakarta (2024) and Mogadishu (2025) exemplify IoT’s practical potential in urban flood management and smart drainage infrastructure. The findings emphasise that integrating IoT into disaster management enhances preparedness, minimises response delays, and fosters sustainable resilience against natural and anthropogenic threats.