Granger Causality-Based Forecasting Model for Rainfall at Ratnapura Area, Sri Lanka: A Deep Learning Approach

Tanzania heavily depends on United States Dollar (USD) foreign currency to import various goods and services into the country. Failure to correctly forecast exchange rates between USD and Tanzanian Shillings (TZS) may pose risks such as inability to import intended goods and services, possibility of losing money in stock exchange markets and other investment businesses in case of unexpected currency appreciation or depreciation as well as poor investment decisions in foreign exchange markets. To address this, this study has developed and comparatively evaluated performances of LSTM (Long Short-Term Memory) and BiLSTM (Bidirectional LSTM) deep learning models for forecasting daily USD to TZS exchange rates. The findings reveal that, BiLSTM model outperforms LSTM in forecasting daily USD to TZS exchange rates, achieving a MAPE (Mean Absolute Percentage Error) score of 0.363 on test set (unseen data) compared to a MAPE score of 1.471 achieved by LSTM model. This study recommends to the prospective Artificial Intelligence (AI) researchers and software developers to use BiLSTM instead of LSTM model to forecast (predict) USD to TZS exchange rates. Also, this study has developed USD to TZS exchange rates dataset which can be used by AI researchers, saving them time and costs involved with creating datasets from scratch. This study has also developed ready to use BiLSTM and LSTM models which can be used by Tanzanian business men and women involved in stock exchange markets, foreign exchange markets and other businesses, to predict daily USD to TZS exchange rates and make appropriate business and investment decisions.

Abstract. The information available on the internet nowadays is diverse and moves very quickly. Information is becoming easier to obtain by the general public with the numerous online media outlets, including news portals that provide up-to-date information insights. Various news portals earn revenue from advertising using pay-per-click methods that encourage article writers to use clickbait techniques to attract visitors. However, the negative effects of clickbait include a decrease in journalism quality and the spread of hoaxes. This problem can be prevented by using text classification to classify clickbait in news titles. One method that can be used for text classification is a neural network. Artificial neural networks use algorithms that can independently adjust input coefficient weights. This makes this algorithm highly effective for modeling non-linear statistical data. The artificial neural network algorithm, especially the Long Short-Term Memory (LSTM), has been widely used in various natural language processing fields with satisfying results, including text classification. To improve the performance of the neural network model, adjustments can be made to the model's hyperparameters. Hyperparameters are parameters that cannot be obtained through data and must be defined before the training process. In this research, the Long Short-Term Memory (LSTM) model was used in clickbait classification in news titles. Sixteen neural network models were trained with different hyperparameter configurations for each model. Hyperparameter tuning was carried out using the random search algorithm. The dataset used was the CLICK-ID dataset published by William & Sari, 2020[1], with a total of 15,000 annotated data. The research results show that the developed LSTM model has a validation accuracy of 0.8030, higher than William & Sari's research, and a validation loss of 0.4876. Using this model, researchers were able to classify clickbait in news titles with fairly good accuracy. Purpose: The study was to develop and evaluate a LSTM model with hyperparameter tuning for clickbait classification on news headlines. The thesis also aims to compare the performance of simple LSTM and bidirectional LSTM for this task. Methods: This study uses CLICK-ID dataset and applies different text preprocessing techniques. The dataset later was used to build and train 16 LSTM models with different hyperparameters and evaluates them using validation accuracy and loss. This study uses random search for hyperparameter tuning. Result: The results of the study show that the best model for clickbait classification on news headlines is a bidirectional LSTM model with one layer, 64 units, 0.2 dropout rate, and 0.001 learning rate. This model achieves a validation accuracy of 0.8030 and a validation loss of 0.4876. The results also show that hyperparameter tuning using random search can improve the performance of the LSTM models by avoiding zero probabilities and finding the optimal values for the hyperparameters. Novelty: This study compares and analyzes the different preprocessing methods on text and the different configurations of the models to find the best model for clickbait classification on news headlines. The study also uses hyperparameter tuning to tune the model into the best model and finding the optimal values for the hyperparameters.

The developed research includes an Long Short-Term Memory (LSTM) network with IoT sensor data for the management of rice crop irrigation. Optimizing the amount of water required for rice crops is very critical, based on real-time predictions of the actual environmental and soil parameters including moisture, temperature, humidity, and rainfall as well as all other possible related factors. IoT sensors installed in agricultural fields gathered data, which was transmitted to a controller and then uploaded to the cloud for further processing. The training dataset contained 5,600 records. In this research, 70% was used for training and the rest 30% for testing. Three LSTM models were used for analyzing sensor data and predicting the need for irrigation and controlling the irrigation pump accordingly. These were Vanilla LSTM, Stacked LSTM, and Bidirectional LSTM. It was evaluated the performance of the models developed using precision, recall, F1 score, accuracy, and other performance measures such as Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> Score. Bidirectional LSTM was determined to have the best performance, with MAE at 0.15, RMSE at 0.24, and an R2 score of 0.93, exhibiting high accuracy and predictive power. The Stacked LSTM and Vanilla LSTM also gave fair results, in terms of having R2 values of 0.89 and 0.84, respectively. The number of training and memory used for the computation were also given, with Vanilla LSTM taking least, followed by Stacked LSTM and Bidirectional LSTM. These results suggest that the LSTM networks can potentially be capable of improving the irrigation management such that more water is conserved and, subsequently, more suitable for agricultural productivity.

This study examines the efficacy of three distinct machine learning models for sentiment analysis: Bidirectional Long Short-Term Memory (Bi-LSTM), Long Short-Term Memory (LSTM), and Logistic Regression. One of the most important tasks in natural language processing is sentiment analysis, which entails categorizing reviews as positive or negative. Fifty thousand sentiment-labeled movie reviews make up the dataset. To determine the most precise and effective technique for sentiment categorization, we put these models into practice and evaluated their respective performances. The highest accuracy of 89.42% is achieved by Logistic Regression, with precision of 88.35%, recall of 91.01%, and F1-score of 89.66%. The LSTM model, well-known for capturing temporal dependencies in text, obtained an accuracy of 86.23%, precision of 89.60%, recall of 82.22%, and an F1-score of 85.75%. The Bi-LSTM model, which processes input sequences in both forward and backward directions to better capture context, demonstrated accuracy of 87.65%, precision of 88.67%, recall of 86.54%, and an F1-score of 87.60%. The results show that although Logistic Regression performs better in accuracy, the Bi-LSTM model offers a substantial trade-off between precision and recall, making it a viable option for sentiment analysis applications. Despite its minor accuracy lag, the LSTM model provides important insights into the sequential relationships of the text data.

This study compares the performance of Long Short-Term Memory (LSTM) and Bidirectional LSTM (Bi-LSTM) models in predicting earthquake occurrences in the Tokai region, using data from the United States Geological Survey (USGS) dataset. Given the importance of accurate earthquake prediction, particularly in high-risk regions, this research focuses on assessing the effectiveness of each model in identifying occurrence and non-occurrence events. Both models were tuned to optimize sensitivity and specificity through adjustments in sequence length, learning rate, and additional hyperparameters, with results evaluated using metrics including sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and area under the curve (AUC). Findings reveal that while both models achieved high sensitivity, the LSTM model demonstrated superior specificity and AUC, indicating a more balanced performance in distinguishing between earthquake occurrences and non-occurrences. The results show that LSTM outperforms Bi-LSTM in terms its classification metrics. LSTM achieved an accuracy of 76%, compared to 55% for Bi-LSTM. For the AUC metric, LSTM scored 66%, while Bi-LSTM scored 67%.

This study addresses the accuracy challenges in e-commerce sentiment classification and thus provides valuable insight for businesses to enrich strategies toward the interpretation of customer feedback and improvement of product development. This article elaborately contrasts long short-term memory (LSTM)-based models with traditional machine learning models, like support vector machines (SVM), random forest, and Naive Bayes classifiers. The authors have used a large dataset of customer reviews from famous e-commerce websites, pre-processed for noise reduction and standardization of the input. Our LSTM models were implemented using state-of-the-art deep learning frameworks, with special consideration while performing hyperparameter tuning. The results showed that the bidirectional LSTM model had the best performance of 92.1% in accuracy, 91.8% in precision, 92.0% in recall, and an F1-score of 91.9%. In comparison, traditional machine learning approaches gave accuracies: 86.5% for SVM, 85.1% for random forest, and 82% for Naive Bayes. Hyperparameter tuning returned the best configurations for the LSTM models: 128 LSTM units, a dropout of 0.3, with learning rates 0.001 for LSTM and 0.0001 for bidirectional LSTM. These optimizations contributed hugely to the performance improvements observed in these models. Error analysis gave insights into the challenges that the models faced. Sarcasm and irony accounted for 22% of the classification errors, while mixed sentiment accounted for 18%, and implicit accounted for 15%. To sum up, this research has shown the efficiency of LSTM models on e-commerce user review sentiment analysis, especially bidirectional LSTM. These models show a much better result in sentiment classification compared to traditional machine learning techniques, which proves them to have huge potential for improving the accuracy of the task in real-world applications.

Wearable electroencephalography (EEG) devices for seizure detection accuracy and reliability are deep learning (DL) applications in the field of epilepsy diagnosis. In this study, we sought to increase the accuracy of seizure detection using advanced DL algorithms on the Children’s Hospital Boston - Massachusetts Institute of Technology (CHB-MIT) EEG database. First, a fully convolutional network (FCN) was trained and assessed using accuracy and recall/precision metrics, and the early stopping technique was used to avoid overfitting. To assess the performance, the FCN was evaluated in terms of various metrics, including accuracy, precision, recall, F1-score, and receiver operating characteristic (ROC)-area under the curve (AUC). In addition, two-dimensional (2D) convolutional neural networks (CNNs) and long short-term memory (LSTM) models were used to model the database, and their performance was thoroughly measured using different metrics, graphs, and confusion matrices. Using LSTM variants, such as standard LSTM, bidirectional LSTM, stacked LSTM, and LSTM attention mechanisms, hybrid convolutional LSTM (ConvLSTM) models were trained and compared. The comparison was conducted based on the training and validation accuracy and loss, as well as the graphs resulting from the precision–recall curves. Apart from DL approaches, EEG signal analysis using time–frequency techniques, such as wavelet transform and short-time Fourier transform, has also been investigated. These methods assisted in the analysis of the time–frequency features of EEG signals in combination with DL models. This study demonstrates that the performance of wearable EEG devices can be augmented using a combination of DL and seizure signal processing techniques. The FCN achieved an accuracy of 92%, with a recall for seizures of 33%, an F1-score of 0.03, and strong ROC-AUC results. The 2D CNN achieved 96% accuracy, a seizure recall of 70%, an F1-score of 0.12, and an ROC-AUC score of 78%. The baseline LSTM struggled with effectiveness at 53% accuracy with a seizure recall of 18%. In contrast, the LSTM model, which incorporated synthetic minority oversampling technique (SMOTE) balancing, was able to reach improvements of up to 89% accuracy, with a precision of 91%, a recall of 86%, an F1-score of 0.89, and a strong ROC curve performance. Among the models, the LSTM with SMOTE was the best performer, with 89% accuracy, 91% precision, 86% recall, and an F1-score of 0.89. These results provide evidence that applying techniques for data balancing in combination with certain DL network architectures significantly improves the detection of seizures using wearable EEG devices worn on the body. We believe that real-time monitoring and high-performance systems are feasible using optimized DL frameworks. The analysis of the performance of different models allows for the understanding of the possibilities of optimizing the architectures of DL algorithms for the modern diagnosis of epilepsy in real time. The source code used to carry out the experiments is publicly available at CHB-MIT EEG Dataset Python Scripts (https://www.kaggle.com/code/adnankust/adnaneeg1).

The system proposed in this paper aims to predict cryptocurrency prices using Bi-Directional Long Short- Term Memory (LSTM), leveraging historical data obtained from Yahoo Finance and CoinGecko APIs. The goal is to assess LSTM models effectiveness in forecasting cryptocurrency prices and offer an interactive interface for users to visualize historical and forecasted prices. Several research works have been conducted on the prediction of cryptocurrency prices through various Deep Learning (DL) based algorithms. This project comprises two main approaches : one involves data analysis, LSTM modeling, and change point detection using Yahoo Finance data, while the other focuses on LSTM model training and price prediction using CoinGecko API data. The paper suggests that the prediction models it presents are useful for traders, investors, [6] and finance academics and are close to accurate at predicting the values of cryptocurrencies. Future research will examine more advanced deep learning architectures, primarily Transformer-based models like the GPT series, to improve pattern detection in bitcoin data. Integrating other data sources, such as sentiment analysis or blockchain measurements, may increase the accuracy of forecasting. With further research into cutting-edge techniques, cryptocurrency forecasting will get better and provide stakeholders with more information to help them make informed decisions. Keywords— Cryptocurrency ; forecasting ; Bi-Directional LSTM Model ; Time-series forecasting ; Machine learning

The objective of this research is to develop a Deep Learning model to forecast the stock price, by using the variant of Long Short-Term Memory. This model predicts the close price of the stock for the future selected date, choosing as inputs the following data: open, high, low, adj close and close prices. This model shows a comparative analysis between three different LSTM networks: Long Short-Term Memory (LSTM), Stacked Long Short-Term Memory (Stacked LSTM), and Stacked Bi-directional Long Short-Term Memory (Stacked Bidirectional LSTM) concluding which one is the best and implementing the model using that variant. We have used the historical stock prices data from yahoo’s financial website over 5 years, by choosing multiple datasets: Apple, Amazon, Google, Meta, Microsoft and Tesla (daily values). In order to get effective in the forecasting model, we have tested the network with different iterations and epochs. The model represents Multiple Graphs for data visualization in different comparisons. We have estimated the effectiveness of our proposed model by using the following performance indicators: the Mean Square Error (MSE), the Root Mean Square Error (RMSE), and the R-Squared of the model. The experimental results clearly show that our Stacked Bi-LSTM model has the highest accuracy values when comparing with the LSTM and Stacked LSTM Models. Hence, we can conclude that our Stacked Bi-LSTM Model is suitable for accurate prediction of the stock market time series. Key Words: stock price prediction, Machine Learning, stacked LSTM, Bi-directional LSTM, Deep Learning, Data pre-processing techniques, Data normalization, Data Visualization, Training and Testing Set, Financial Time Series, Future prediction

Phishing attacks continue to pose a serious threat to cybersecurity, underscoring the need for effective and scalable detection methods. This study evaluates the performance of Recurrent Neural Network (RNN) architectures—specifically Long Short-Term Memory (LSTM) and Bidirectional LSTM (BiLSTM)—for detecting phishing websites based on the sequential patterns in URL structures and webpage content. The LSTM model achieved an Area Under the Curve (AUC) of 0.92, with an overall accuracy of 98.4%, precision of 98.9%, and recall of 97.1%. These results indicate a strong ability to identify phishing URLs with a low false positive rate, although performance declined when detecting sophisticated or zero-day phishing attempts. The BiLSTM model, which incorporates bidirectional context, achieved a higher AUC of 0.95 and improved precision of 91% at a recall of 89%. However, it exhibited a slightly lower overall accuracy of 97.9% and a higher false negative rate. Both models effectively differentiated phishing from legitimate URLs, with BiLSTM offering improved context awareness but at the cost of reduced recall. The results suggest that while BiLSTM enhances contextual understanding, the LSTM model offers better generalization and computational efficiency for real-time deployment. This work highlights the potential of RNN-based models in phishing detection and the importance of balancing sensitivity and specificity in cybersecurity applications.

Surface Temperature (ST) is important in terms of surface energy and terrestrial water balances affecting urban ecosystems. In this study, to process the nonlinear changes of climatological variables by leveraging the distinct advantages of Long Short-Term Memory (LSTM) and Bidirectional Long Short-Term Memory (BiLSTM), we propose an LSTM-BiLSTM hybrid deep learning model which extracts multi-dimension features of inputs, i.e., backward (future to past) or forward (past to future) to predict ST. This study assessed the climatological variables, i.e., wind speed, wind direction, relative humidity, dew point temperature, and atmospheric pressure impact on ST using five major coastal cities of India: Chennai, Mangalore, Visakhapatnam, Cuddalore, and Cochin. The Recurrent Neural Networks (RNN) and hybrid LSTM-BiLSTM models have effectively predicted ST and outperformed the standalone Artificial Neural Networks (ANN), LSTM, and BiLSTM models. The RNN and LSTM-BiLSTM models have performed better in predicting ST for Mangalore (Nash-Sutcliffe efficiency (NSE)=0.91), followed by Cochin (NSE=0.89), Chennai (NSE=0.88), Cuddalore (NSE=0.88), and Vishakhapatnam (NSE=0.81). The hybrid data-driven modeling framework indicated that coupling the LSTM and BiLSTM models was proven effective in predicting the ST of coastal cities.

With the advancement of cloud computing technologies, there is an ever-increasing demand for the maximum utilization of cloud resources. It increases the computing power consumption of the cloud’s systems. Consolidation of cloud’s Virtual Machines (VMs) provides a pragmatic approach to reduce the energy consumption of cloud Data Centers (DC). Effective VM consolidation and VM migration without breaching Service Level Agreement (SLA) can be attained by taking proactive decisions based on cloud’s future workload prediction. Effective task scheduling, another major issue of cloud computing also relies on accurate forecasting of resource usage. Cloud workload traces exhibit both periodic and non-periodic patterns with the sudden peak of load. As a result, it is very challenging for the prediction models to precisely forecast future workload. This prompted us to propose a hybrid Recurrent Neural Network (RNN) based prediction model named BHyPreC. BHyPreC architecture includes Bidirectional Long Short-Term Memory (Bi-LSTM) on top of the stacked Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU). Here, BHyPreC is used to predict future CPU usage workload of cloud’s VM. Our proposed model enhances the non-linear data analysis capability of Bi-LSTM, LSTM, and GRU models separately and demonstrates better accuracy compared to other statistical models. The effect of variation of historical window size and training-testing data size on these models is observed. The experimental result shows that our model gives higher accuracy and performs better in comparison to Autoregressive Integrated Moving Average (ARIMA), LSTM, GRU, and Bi-LSTM model for both short-term ahead and long-term ahead prediction.

In recent years, research and development of brain-computer interfaces (BCIs) have progressed rapidly, aiming to support communication for patients with quadriplegia resulting from conditions such as amyotrophic lateral sclerosis (ALS) or neurological damage. However, the classification of electroencephalogram (EEG) signals remains a significant challenge, with many studies limited to binary classification and insufficient accuracy for practical application. This study aims to classify Japanese vowels during covert speech imagery using deep learning techniques based on brain activity signals. EEG data were recorded from a healthy adult male subject during imagined articulation of Japanese vowels without vocalization. Support Vector Machine (SVM), Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and Bidirectional LSTM (BiLSTM) models were trained and evaluated on the collected data. The classification accuracies achieved were 65%, 90%, 85%, and 93% for SVM, CNN, LSTM, and BiLSTM, respectively. These results indicate that deep learning approaches can successfully identify Japanese vowels from EEG signals associated with covert speech imagery. Future work will focus on applying natural language processing techniques to the predicted character sequences to enable word- and sentence-level estimation, ultimately contributing to the development of a communication support system for non-vocal environments.

Many countries around the globe, including Tanzania, are investing in renewable energy, partly because of the need to increase electric power generation and meet the ever-increasing electricity demands, and partly because of the need to combat global climate change caused by carbon emissions from fossil fuel-based electricity production. Kishapu Solar Project, located in Kishapu District, Shinyanga Region, Tanzania, is one of the renewable energy projects with a planned generation of 150 MW of electric power once completed, using solar photovoltaic technology. Accurate forecasting of solar radiation is important in managing and maintaining solar energy plants. Inaccurate forecasting of solar radiation can lead to poor harvest of solar energy and poor management of solar energy plants, which directly impacts electricity contribution to the national electricity grid. Long Short-Term Memory (LSTM) and its variant, PeepHole LSTM, are examples of Deep Learning models that can be used to predict solar radiation. The purpose of PeepHole LSTM is to introduce “peephole” connections in the LSTM model, which allow LSTM gates to directly inspect memory cell state and hence have richer information to process. However, there is still a research gap on the impact of “peephole” connections on the prediction performance of the LSTM model, especially in the context of forecasting solar radiation in Kishapu District, where the Kishapu Solar Project is located. The purpose of this study is to fill this research gap by developing both LSTM and PeepHole LSTM models and comparing their performances in forecasting solar radiation in Kishapu District. The 14-year Kishapu District daily solar radiation (Downward Shortwave Radiation) dataset was first pre-processed by scaling (normalizing) it using Standard Scaler and then split into training set (the first 70%), validation set (the next 15%) and test set (the last 15%) before being utilized to train and test the developed LSTM and PeepHole LSTM models. Both LSTM and PeepHole LSTM models used the previous 5 days' solar radiation values as inputs to predict the output (next day (6th day) solar radiation value). The trained and better-performing model between the two was finally saved in .h5 format and integrated with a Gradio based Web Application to provide a user interface for officials in Kishapu District and at Kishapu Solar Project to predict daily solar radiation. The findings from performance evaluation revealed that the LSTM model outperformed the PeepHole LSTM model in predicting solar radiation by achieving test Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) scores of 26.22 W/m2 and 36.65 W/m2, respectively, compared to MAE and RMSE scores of 30.84 W/m2 and 41.88 W/m2, respectively, achieved by the PeepHole LSTM model, concluding that “peephole” connections reduce the performance of the LSTM model rather than increasing it when used to forecast solar radiation in Kishapu District. This study’s recommendation to Artificial Intelligence (AI) software developers and researchers is to use the LSTM rather than the PeepHole LSTM Deep Learning model when forecasting solar radiation in Kishapu District and in locations with similar climatic conditions

Most of the current word segmentation methods are rule-based and traditional machine learning methods. Universal word segmentation tools do not work well in the field such as metallurgy. Domain-specific Chinese word segmentation is rarely studied. In recent years, with the development of deep learning, the neural network has been proved to be effective in Chinese word segmentation. However, this promising performance relies on large-scale training data. Neural networks with conventional architectures cannot achieve the desired results in low-resource datasets due to the lack of labeled training data. This paper takes the field of metallurgy as an example and proposes a domain-specific Chinese word segmentation based on Bi-directional long-short term memory (Bi-directional LSTM) model in the metallurgical field. First, the word segmentation model is obtained by using the Bi-directional LSTM model to train the internal and external domain knowledge. Then, a series of tuning parameters are carried out and the label probability of the word is combined with the weight. Finally, the result of word segmentation is obtained by label inference layer. The experimental results show that the proposed method can create a better word segmentation effect in the field of metallurgy.