Sarbagya Ratna Shakya

Social media analytics has emerged as a critical tool for investigations, providing valuable insights into user interactions, trends, and behaviors across various platforms. With over 4.76 billion active social media users globally, this vast data source has proven useful in government investigations and criminal litigations. However, extracting reliable information from social media presents significant challenges, including data representativeness, ethical and legal concerns, user behavior interpretation, and the growing issue of misinformation. This paper reviews recent trends in social media analytics, particularly its application in investigations, highlighting the increasing use of machine learning for data processing and analysis. Our analysis reveals its transformative impact, enabling advanced evidence collection, suspect identification, criminal network mapping, and proactive crime prevention. However, significant challenges persist regarding data validity, bias minimization, ethical responsibilities, and misinformation. Additionally, the paper offers future research directions, emphasizing the need for standardized frameworks, adaptive ethical guidelines, AI-driven real-time multimodal analysis, and robust privacy protection measures. These advancements aim to enhance the reliability and effectiveness of social media analytics for investigative purposes.

Since access to data has been on the rise and computational capabilities have improved, the prediction of stock prices has become very important. The work in this paper aims at integrating sentiment analysis and historical stock data to improve the accuracy of stock price forecasting. In our study, we have used Sentiment scores, which were derived from tweets and news using a zero-shot transformer-based model and combined with traditional stock metrics to predict future stock prices. To evaluate the accuracy of the models, three regression models based on recurrent neural networks: LSTM, BiLSTM and CNN+BiLSTM were compared with each other for different combination of input variables. Predictive accuracy was evaluated using statistical performance indicators including R-squared values and mean squared error. The results indicate that all the models could identify the trends of the stock movements. However, the BiLSTM model provided the best results and outperformed the other two models in terms of accuracy. The study demonstrated that incorporating sentiment scores significantly improves stock price prediction accuracy and offers a more comprehensive approach to stock prediction. Also, this study suggests the appropriate combination of input variables, such as sentiment and historical data, and the right model that can give the best stock price prediction. This work shows how qualitative sentiment data can be combined with quantitative stock metrics to enhance the predictive capability and may be helpful to investors and traders.

The Internet of Things (IoT), which was first conceptualized in 1999 [1], has a crucial application in the industry. The Industrial Internet of Things (IIoT) refers to the interrelated, automated use of sensors, devices, and machines that run on industrial applications to increase efficiency and reliability of manufacturing. It provides connectivity between systems, machines, and people in companies. Many companies need to improve asset management and maintenance by reducing waste and cost to improve their competitive advantage. IIoT helps to increase operational efficiency in the manufacturing industry by integrating information technology with operational technology [2]. IIoT is also commonly known as Industry 4.0, which works on interconnected smart machinery, embedding sensing devices, data analytics, and automated decision-making with edge computing technology to increase productivity, make supply chain efficient, and improve distribution, capacity, resource management [3], worker safety, and return on investment [4]. IIoT integrates technology related to Artificial Intelligence (AI) with manufacturing learning and responding, analyzing big data in operations and productions [5]. This integration improves manufacturing performance, scalability, and ideas behind evolving and improving the business by capturing data from sensors and communicating accurately and consistently [6]. The raw data collected from these sensors and assets consist of the digital footprints of the manufacturing product and process. These data, when analyzed, can provide information regarding the production and supply chain that can be implemented in decision-making and optimization. Although IIoT has added value to business operations, the adoption of these digital technologies has still been difficult for businesses. In emerging economies, the slow and low implementation has slowed the progress in these fields [7]. Also, interconnecting different sensors, actuators, and controllers with the production lines and equipment to perform the manufacturing process automatically without any or minimum human intervention has added challenges in terms of safety and security of industrial production and automation [8]. Because of this reason, some companies have shown resistance and reservations in adopting these technologies. However, with the improvement of these technologies, many manufacturers are now switching to implementing IIoT in their manufacturing process.

Diabetes is a chronic condition that affects how the body processes blood sugar. Early diagnosis and management of diabetes are essential for preventing these complications. Machine Learning (ML) techniques offer an effective means to accurately diagnose diabetes by identifying key risk factors and developing predictive models. In this study, we assess the performance of 11 ML algorithms on four diabetes prediction datasets, considering the top 2, top 3, and all attributes. Through k-fold cross-validation, we ensure robust and generalizable results. We use a set of standard evaluation metrics such as accuracy, precision, recall, f1-score, and Receiver Operating Characteristic curve (ROC_AUC). Our analysis aims to determine the optimal number of features and assess how performance changes with feature additions. Notably, some ML classifiers achieve satisfactory classification and predictive abilities using only the top 2 or 3 features. Furthermore, varying dataset performances across algorithms highlight the need for assessing multiple models to identify the most suitable one. These findings enable the creation of dependable models that enhance patient outcomes by leveraging effective algorithms and pertinent features.

The Housing price index (HPI) plays a key role in real estate dynamics, which majorly contributes to the United Sstates economy. Forecasting accurate housing price indices will be very beneficial for investors, policymakers, banks, and homeowners. There was a wide range of approaches implemented for forecasting HPI using microeconomic features with very little concentration on macroeconomic features such as economic indicators at a national level. HPI directly correlates with these macroeconomic features, which have a direct influence on overall real estate dynamics and the economy. In this paper, we applied various time series, machine learning [ML], and deep learning [DL] models to forecast HPI using macroeconomic features. Our goal is to get accurate predictions and get the relationships that directly impact the housing market. Macroeconomic features of the dataset were collected from economic and federal agencies with the aim of forecasting HPI values. We implemented the autoregressive model integrated moving average model [ARIMA] and Facebook Prophet [FB Prophet] to know the temporal dependencies. Next, we applied machine learning and deep learning models Random Forest, Extreme Gradient Boosting (XGBoost), Long Shorth term model (LSTM), Recurrent neural networks (RNN), and a hybrid approach, which is an ensemble of Extreme Gradient Boosting and LSTM. We studied the efficacy of implemented models and evaluated the performance of the models using the performance metrics Mean Absolute Error (MAE), and Root Mean square Error (RMSE), and R-squared.

With the increasing demand for smarter antenna design in advanced technology applications, well-designed antennas have been an important factor in enhancing system performance. Most traditional antenna design requires multiple iterations and extensive testing to produce a final product. Machine learning (ML) algorithms have been used as an alternative to predict the optimal design parameters, but the outcome depends highly on the ML model efficiency. With recent development in machine learning algorithms and the availability of data for antenna design, we investigated different machine learning algorithms for optimizing the output strength of three basic antennae by analyzing the signal strength of the antenna for various antenna parameters. Different regression-based ML models were used to learn the behaviors and efficiency of three different antennas and to predict the output strength (S11) for different ranges of frequencies. The experiment compared and analyzed these ML regression algorithms for three different antennas: shot antenna, patch antenna, and bowtie antenna. In addition, the paper also provides comparison of ensemble ML models for performance analysis using the best three ML algorithms from the preliminary study. This study optimizes antenna parameters and quicker and smarter antenna design procedure using ML algorithms as compared to traditional design methods.