International Journal of Innovations in Science & Technology
Not a member yet
813 research outputs found
Sort by
Seismic Data Analysis and Earthquake Prediction with IoT Sensors and SmartGRU Model
Full text linkTectonic plate movement causes a slow accumulation of stress in the Earth’s lithosphere, especially around plate borders, leading to earthquakes. An earthquake occurs when this stress overcomes friction along a fault or exceeds the strength of the surrounding rock. Accurate earthquake prediction remains challenging due to the complexity of seismic data and the limitations of traditional methods. This creates a pressing need for models capable of real-time analysis and high prediction accuracy. The Internet of Things (IoT) provides a novel method for detecting earthquakes using a variety of sensors to collect vital seismic data, such as latitude, longitude, depth, magnitude, and time. IoT controllers and centralized systems process and analyze this data to enable efficient monitoring and forecasting. Furthermore, with the help of a machine learning model named Bidirectional Gated Recurrent Unit (Bi-GRU), which integrates sophisticated data fusion and advanced machine learning techniques. Our proposed study model, SmartGRU, demonstrates how to improve earthquake prediction systems by combining IoT sensors with a Bi-GRU machine learning model that incorporates an emerging approach
A Framework for Fraud Detection in Banking Transactions Using Machine Learning and Federated Learning
Full text linkThe digital banking revolution has transformed financial services to make payment faster, more convenient, and borderless. But with this revolution came an abrupt increase in fraudulent transactions through credit cards that threatening both the financial institutions and the customers. While conventional fraud detection mechanisms are not capable of addressing new-generation fraud patterns, there is an increasing demand for intelligent, adaptive, and secure solutions with high precision without any data privacy compromise. Proposed model leverages four machine learning models, Linear Regression, Decision Tree, Long Short-Term Memory (LSTM) and Convolutional Neural Network (CNN). LSTM and CNN are used due to their power in learning complicated sequential and feature-based patterns, with Decision Tree and Linear Regression added due to their ease, quick execution, and interpretability. Every model is locally trained on partitioned banking datasets for each simulated client. Model parameters are combined with the Federated Averaging (FedAvg) algorithm to create a globally shared fraud detection system. Experimental testing was conducted on a real-world banking transaction data set published in a non-IID manner to mimic real-world client situations. The federated learning paradigm achieved encouraging results: CNN and LSTM models achieved detection accuracy rates of over 95%, with outstanding performance in the detection of hidden or time-series-based fraud patterns. The Decision Tree model also achieved steady performance at 91% accuracy, and Linear Regression achieved a reasonable baseline at 88%. These results indicate that even simple models, when used in a collaborative federated environment, can contribute meaningfully to fraud detection. This research contributes to the body of research supporting federated banking solutions and fills a significant gap by demonstrating how several ML models can coexist and collaborate in a decentralized setup for fraud detection through credit card transactions
Micro Hydro Power in Pakistan: A Comprehensive Review of Development, Applications, Challenges, and Future Prospects
Full text linkAmid Pakistan’s evolving energy landscape—marked by a 62.1% fossil fuel dependency and persistent rural-urban access disparities—micro hydro power (MHP) systems offer a cost-effective and environmentally resilient solution. This review integrates technical, economic, and policy perspectives to evaluate the current and potential role of MHP in the country. It presents a structured analysis of turbine technologies (Pelton, Cross-flow, Kaplan, Turgo), their performance characteristics, and appropriate deployment contexts across various head and flow conditions. Drawing on case studies from Khyber Pakhtunkhwa, Gilgit-Baltistan, and Punjab, the study highlights site-specific generation capacities, operational challenges (e.g., sedimentation, seasonal variability), and socio-economic impacts. Furthermore, it explores the institutional governance structure WAPDA, AEDB, PEPCO, and IPPs—and national policy initiatives under CPEC and the Alternative Energy Policy. The findings reveal that Pakistan’s mini-hydro capacity remains underutilized despite recent advancements in turbine efficiency and feed-in tariffs. Strategic expansion of MHP through modular designs, smart grid integration, and rural electrification incentives could significantly bridge the country’s energy access gap while aligning with its 2030 renewable energy targets.
 
A Signal-Decomposed Ensemble Forecasting and Classification Framework for Household Power Consumption: An STL-Inspired Machine Learning Approach
Full text linkAccurate short-term forecasting of residential power consumption is crucial for smart grid stability, real-time energy optimization, and personalized demand-side management. Traditional time-series and standalone AI models often struggle with the nonlinear, nonstationary, and noise-sensitive nature of high-resolution household load data. Unlike existing models, this study introduces an STL-based residual decomposition fused with lag-aware ML forecasting and threshold-based classification under real-world conditions. To address these challenges, this study proposes a novel STL-inspired decomposition framework integrated with four machine learning models, i.e., Least Squares Boosting (LSBoost), Bagging, Support Vector Regression (SVR), and Multilayer Perceptron (MLP), for forecasting and classification of normalized household energy consumption. The methodology begins with robust preprocessing, including IQR-based outlier removal and min-max normalization, followed by STL-like decomposition into trend, seasonal, and residual components. Lag-based features from the residual signal are used for forecasting via the selected ML regressors. Final predictions are reconstructed and threshold-classified into OK/NOT OK categories to simulate alert-based power decision scenarios. Experimental validation on the UCI Household Power Consumption dataset reveals that SVR achieves the best trade-off among all models, with RMSE = 0.0267, MAE = 0.0193, MAPE = 12.5%, and Pearson correlation coefficient = 0.846. For classification performance, SVR also attains an AUC of 0.941 and a binary classification accuracy of 93.7%. The synergy between STL decomposition and residual-based modeling not only improves regression accuracy but also facilitates threshold-aware classification with high interpretability. Additional visual diagnostics including forecast overlays, residual histograms, ROC curves, and Q–Q plots demonstrate the model’s interpretability and robustness. The proposed ensemble framework not only enhances prediction accuracy but also ensures practical deployment feasibility through threshold-aware decision modeling
Extreme Flooding in Pakistan: An AI-Powered Framework for Enhanced Urban Flood Management System
Full text linkUrban flooding poses considerable challenges for metropolitan areas, contributing to rapid urbanization and significant climatic change. This research develops a machine learning-based Urban Flood Management System (UFMS) to predict and manage flood risks, incorporating an enhanced risk warning system for rapidly urbanizing areas. The mitigation of urban flooding parameters, such as rainfall intensity, humidity, temperature, soil moisture, land use, and drainage network capacity, is analyzed in the UFMS. The system employs the artificial intelligence model Support Vector Machine (SVM), in aggregation with ARIMA modeling, to attain a remarkable accuracy rate of 99.99% to forecast flood events. The model undergoes training with two decades of historical meteorological data to augment its predictive prowess and guarantee robust performance. The result shows that SVM performs with superior accuracy in comparison to other machine learning algorithms (MLAs) by effectively handling complex, multidimensional and multimodal data. This hybrid methodology provides real-time and highly accurate prediction of upcoming floods that leads to actionable insights for urban planners and emergency response teams. Future improvements may involve the utilization of real-time data obtained from Internet of Things (IoT) nodes combined with an advanced deep learning model to improve forecast accuracy, scalability and reduce response time, which will lead to minimizing damages
Tracking Temporal Migration of the Indus River: Morphological Changes in a Downstream Reach
Full text linkIndus River morphological changes create environmental challenges, impacting local communities and ecosystems through fertile land loss, bank erosion, and higher flood risks. Monitoring these changes is crucial for flood and water resource management and infrastructure protection. This study uses geospatial data and tools to analyze spatial and temporal morphological dynamics of a downstream Indus River reach between the Sukkur and Kotri barrages from 1995 to 2024. Satellite imagery was analyzed to study morphological changes. Significant channel adjustments in river shape and form were observed, evident through erosion, deposition, and lateral shifts over the past three decades. The maximum erosion, covering 35,540 ha, and accretion, covering 23,737 ha, were observed between 1995 and 2005, with later periods showing reduced erosion and greater stability. The total cumulative erosion was 71,575 ha, and the total cumulative accretion was 64,790 ha, which gives a net loss of 6,785 ha. The sinuosity analysis showed that the meandering tendency of the river increased over the years as the sinuosity ratio increased from 1.82 in 1995 to 1.93 in 2024. These findings reveal the features of fluvial dynamics of the Indus River and stress the importance of reducing the adverse effects of these changes as necessary for the area\u27s sustainable development
Spatio-Temporal Analysis of Meteorological Drought in Lahore (1995–2024)
Full text linkEvaluating urban meteorology is essential for efficient water resource management, especially considering climate change and an increase in Urban population. It helps to understand the severity and scope of drought conditions, which enables improved planning and execution of drought response measures. This research paper examines long-term patterns of rainfall variability and drought situations in Lahore, spanning a period of 30 years (1995- 2024). Monthly rainfall data taken from the UCSB-CHG/CHIRPS dataset, along with potential evapotranspiration (PET) information from the TERRACLIMATE dataset, are being analyzed using the Standardized Precipitation Index (SPI) and Standardized Precipitation Evapotranspiration Index (SPEI). The whole dataset is processed by Google Earth Engine (GEE), with Lahore’s administrative boundaries used to define the Area of Interest (AOI). The analysis recognizes significant annual and spatial variability, with the mean annual Precipitation recorded at 65.35mm. Extreme years included 2021, with 184.77 mm, and 2019, in this year precipitation was only recorded at 13.06 mm, which highlights growing climatic inconsistencies. SPI values dipped as low as -2.6 in 2015 in the southern part of Lahore, indicating severe drought conditions, while northern Lahore experienced values as high as 1.7, denoting extreme wetness. SPEI values exhibited a similar pattern, with the southern region recorded -2.3 in 2024, reflecting ongoing moisture stress, contrasted by northern Lahore reaching 1.2 to 2, a marked improvement in hydrological balance. These results show that Lahore is becoming more and more vulnerable to both drought and flooding because of urban growth and changes in the monsoon. According to the findings, localized, data-driven climate adaptation policies that prioritize drought resistance, water conservation, and efficient urban planning are essential
Improving Communication Quality Through Anonymous Communication: An Experimental Study using ANONI Application
Full text linkQuality of communication is closely related to quality education, United Nations Sustainable Goal. The performance of students can be analyzed through effective communication between students with their instructors. Multiple reasons cause poor or ineffective communication from students, for instance, shyness, fear of being criticized, and nervousness. Communication in anonymous mode is explored by various research studies. It is noticeable from the literature that students’ participation is directly related to anonymous communicationincrease. Although anonymous communication has a positive effect on student participation, this anonymous factor also causes disruptions or unanticipated negative intrusions during class discussions. This study aims to improve the quality of anonymous communication and explore the impact of anonymous communication on students with less participation. The study\u27s objective has been achieved about undergraduates enrolled in software engineering programs. The reward-based synchronous & asynchronous web application named “ANONI” was utilized for this purpose. The results show a positive increase in participation and constructive communication of students during the session, as only 2 off-task activities were observed
On The Crosstalk of Circadian Rhythm and Th17 cells: An Integrated Biological Regulatory Pathway
Full text linkTh17 cells play a pivotal role in cell-mediated immunity and also have implications for autoimmune disorders. The interplay between the circadian rhythm and the immune system has driven interest in developing novel therapies. Th17 cells have a robust relationship with the circadian rhythm through clock-controlled genes such as NFIL3 (E4BP4), RORA, RORB, NR3C1, and RORC. The purpose of this study is to construct a literature-curated updated biological regulatory network (BRN) of the molecular regulators of circadian rhythm and CD4+ Th17 cells. The integrated BRN will provide a holistic view of the differentiation process of Th17 cells from a circadian rhythm perspective, which will enhance our understanding of the interplay between the two systems. We aim to perform formal modelling and analysis of this BRN using our previously developed approaches to gain system-wide insights into various molecular expression dynamics and identify the significance of biological clocks in immunity in the future. In addition, biological pathway databases are an integral part of omics analytical workflows, and their continuous updates with the latest knowledge are crucial for gaining biological insights from such studies. Therefore, with this additional objective, we have also uploaded this pathway to WikiPathways (Database), to facilitate its use in future studies, which can be accessed via the following URL: https://classic.wikipathways.org/index.php/Pathway:WP5130. To our knowledge, this is the first study to report a literature-curated pathway of comprehensive regulatory interactions and crosstalk between Th17 cell differentiation and circadian genes
DECS: A Deep Learning Approach for EEG Channel Selection in Emotion Classification
Full text linkThe non-stationary nature of Electroencephalogram (EEG) signals often leads to high computational complexity in emotion recognition systems. To address this, we propose a novel framework that integrates optimal channel selection with efficient feature extraction. Our method begins by converting preprocessed EEG signals into two-dimensional spectrograms using a Continuous Wavelet Transform (CWT). These spectrograms are then processed by a GoogLeNet model for deep feature extraction. A key contribution is the Differential Entropy-based Channel Selection (DECS) technique, which identifies and retains the most informative channels. To manage dimensionality, the extracted features are encoded using the Bag-of-Deep-Features (BoDF) method, which employs k-means clustering to create a visual vocabulary and represents features as histograms. Finally, these histogram features are classified using a Support Vector Machine (SVM). Evaluated on the SJTU SEED and DEAP datasets, the proposed model achieves state-of-the-art classification accuracies of 95.1% and 81.1%, respectively, demonstrating its effectiveness and efficiency