Organizations face various challenges when analyzing tabular data. One of the key challenges is the complexity of business data analytics tasks. These tasks involve handling large volumes of data, organizing and structuring it, and extracting valuable insights. Additionally, employees who work in financial analysis often encounter a significant workload on tools such as Excel, SAP, PowerBI, and Tabula. This workload can result in increased effort and time required to analyze and make sense of the data. Organizations must address these challenges to ensure efficient and effective analysis of tabular data. Organizations spend more money to perform business tasks. Nowadays, there are many efficient models in artificial intelligence to perform text, audio, video, and image-based tasks, but no efficient models are available to perform tabular-based tasks specifically. Pandas Python library provides various functionalities and APIs that are useful for business data analysis. This research solved the problem with tabular data using Python Pandas code generation. Here, the researchers used two datasets, each with 50 records. The Large Language Model (LLM) is a supervised Learning Pre-Trained Foundation Model (SSL PFM) category based on text generation. The SSL PFM helps the language models learn the context of language and world knowledge. During this research, models such as LLaMA-2, Falcon, CodeLlama, and Mistral were considered for analysis. Each of these models consists of a Billion parameters. Moreover, Quantization techniques were incorporated to reduce model size, enabling models to load with minimal hardware. After quantizing the model, Parameter Efficient Fine Tuning (PEFT) trains the dataset using only a few model layers; other layers are frozen. Well-known experts with Pandas evaluate the fine-tuned models of chosen language models. Finally, mistral-7B produced prominent results in analyzing business tasks and producing summarized results.

The study comprehensively analysed public sentiment towards ChatGPT, a language processing AI tool, using approximately 135,000 tweets from an initial dataset of 236,275 entries. After cleaning and pre-processing the data using basic NLP techniques, Machine Learning (ML) models—Long Short-Term Memory (LSTM), Decision Tree (DT), and Support Vector Machine (SVM)—were applied to classify the tweets into positive, negative, and neutral categories. The LSTM model demonstrated superior performance across all evaluation metrics—accuracy, precision, recall, and F1 score—for all tweet categories. Word clouds generated for each category revealed that the most frequent words in positive tweets included ‘student,’ 'education,’ 'research,’ and ‘teacher,’ indicating a strong association of ChatGPT with educational and research contexts. Similarly, ‘AI,’ ‘student,’ 'research,’ ‘course,’ and ‘education’ dominated neutral and negative tweets, highlighting prevalent discussion topics. However, the study had limitations, including dependency on a single dataset, potential loss of contextual information during pre-processing, and a focus on English tweets only. Despite these limitations, the study provides valuable insights into the public’s perception of ChatGPT, underscoring its significant association with education and research and the effectiveness of the LSTM model in classifying tweets. The findings suggest that future research should incorporate diverse data sources and languages better to understand global sentiment towards ChatGPT and similar AI tools.

The spread of the COVID-19 pandemic impacts everyone's lives somehow. The adverse impact on educational systems in all regions has a wide-ranging extension effect. A total lockdown creates new challenges for students to study and teachers to properly direct the class, transforming such a physical classroom system into a virtual classroom. Despite the widespread use of online learning in schools throughout the COVID-19 epidemic, there are inadequate facts about the elements that influence student time utilisation with this innovative learning environment in a crisis. This study focuses on many aspects a student goes through during the pandemic. It's vital to understand whether students utilise their time efficiently during the pandemic, especially virtual learning systems. In this study, two machine learning models, namely decision tree and artificial neural network, were created to predict whether a student has utilised their time properly or not. The data set for the model was gathered via online questionnaires, which consist of 19 questions based on various parameters. Then the dataset is preprocessed using different mechanisms before feeding the data to the models. The final results of ANN show better accuracy of 65.91 and a lesser error rate of 34.08 after the comparison between the two models.

Usually the X-ray images can be used to detect fractures and help the physician to provide the appropriate diagnosis. To avoid further injury or causing more harm to the injured area, it is crucial to treat any broken or fractured bones as medical emergencies and seek the necessary treatment without delay. A thorough analysis was carried out to discover the previous studies in detecting the fractures using X-ray images by incorporating different techniques. Especially, as we all know artificial Intelligence and Machine Learning play a vital role in the medical domain in various aspects nowadays. During this mapping study, various fracture detection using image processing techniques were considered. Studies related to the analysis on X-ray images to detect fractures from 2010-2021 presented along with the findings discovered. The related studies extracted from six electronic archives namely Springer Link, IEEE Xplore, ACM Digital Library, Research Gate, and Science Direct. From the studies obtained there were no studies reported yet to classify the type of femoral neck fracture. There are four types of neck fractures that can occur in the femoral. It’s difficult for a doctor or radiographer with less experience to detect the exact type of femoral neck fracture in limited time. Anyhow to expand this study further, to detect the femoral fracture type in advance, previous studies related in detecting the other fractures using X-rays were analyzed. Totally around 200 studies were initially chosen, and 12 papers were shortlisted from the depth analysis. Numerous studies have been done on the identification of bone fractures. There have been attempts to use Artificial Neural Networks (ANN), Support Vector Machine (SVM), and Local Binary Patterns (LBP) approaches frequently. This study would be a promising start for the future scholars to more precisely focus on detecting the type the femoral neck fracture in advance by adapting prevailing Machine and Deep Learning techniques.

Queue scheduling is an important process that is used when processes can be divided into different classes based on the scheduling needs. In general, queue lengths and waiting times can be predicted with the use of a queuing model; especially, the importance of the services provided by any sector or organization for delivering effective services to their consumers highly depends on managing the queues in an effective manner. So, in general, analyzing and evaluating queues is more significant in fulfilling customer needs in a limited time. The applications and tactics developed earlier would not provide satisfactory solutions to solve this problem. But the studies done in the recent past have utilized newer technologies. Especially by incorporating Machine Learning techniques and various queuing models. During this study, a thorough analysis was carried out to discover the different strategies used in optimizing the queues by presenting a Systematic Literature Review (SLR) by examining available research perspective to queue management systems from 2016 to 2022. Initially 200 studies selected using seven electronic repositories and finally selected 14 for this analysis. The study’s findings reveals that most of the studies have used machine learning approaches, utilizing tools like ARENA,SIMIO and adapting various queuing algorithms to solve this problem.

In today's business world, customer turnover is a significant problem. Communications companies aren't exempt from these problems. Retaining consumers is more important than recruiting new ones when it comes to business. Getting new clients is about five times as expensive as keeping old ones in this field. As a result, anticipating client turnover is a huge challenge for almost all organizations. This study focused on analyzing information on around 7000 post-paid subscribers by considering 21 different attributes. Initially, the data was fed into machine learning techniques such k-nearest neighbors, artificial neural networks, etc. In addition, deep neural networks (DNN) have also considered more than one hidden layer. A total of 4284 of the 7234 post-paid customers are considered non-churners, while the remaining 2950 are churners. The long short-term memory networks (LSTM) considered under the DNN produce results far superior to the other techniques, with the highest accuracy rate of 82.46%. Finally, the LSTM method was used to create the final prediction model.

Analysis of Arterial Blood Gas (ABG) is an important investigation to measure oxygenation and blood acid levels. It is crucial in measuring the clinical status and contributes to an efficient and effective healthcare plan. Generally, ABG is applied in the emergency care units (ECU) and intensive care units (ICU). Most of the time, the doctors and nurses have difficulties identifying the type of respiratory failure with the help of ABG test results. So, during this research with the adaption of certain supervised machine learning approaches, namely Extreme Gradient Boosting (XGBoost), Adaptive Boosting (AdaBoost), Catboost, Random Forest, Naïve Bayes, Support Vector Machine (SVM), LightGBM, K-Nearest Neighbors (KNN), Neural Network (NN) and Decision Tree and have been incorporated with the intension of identifying the type of the respiratory failure with the highest accurate technique. To fulfil this purpose, 700 patient test results have been obtained from a public hospital in Sri Lanka. From the results discovered, XGBoost outperformed against all other techniques in identifying the type of respiratory failure with the highest accuracy of 98.65% and the lowest error rate of 1.35%. To ensure whether the XGBoost outperformed against the different percentages of training and testing data, K-fold cross-validation with five folds also has been performed with the dataset. The cross-validation produces results with an accuracy of 98.45% and the lowest error rate of 1.55%. In conclusion, XGBoost has been utilised in developing the prediction model. This would be a promising start for a future research scholar to adopt the hybrid techniques and the deep learning techniques to identify the causes of respiratory failure and the prediction of the type of respiratory failure.

Nowadays, one out of four people above 25 will suffer from a stroke. Especially this year, with the highest count of around 13.7 million people discovered with stroke for the first time. Out of 13.7 million, 5.5 million were fatalities. This was stated in a recent WHO study. It is estimated that if no action is taken, the number of fatalities will rise to 6.7 million yearly. The pandemic situation of COVID-19 will play a significant cause in the expanded death rate of stroke. Even for adults and patients with minor risk factors affected by stroke rather than in previous years. This study predicts the impact level of stroke with the development of an ensemble model by combining the various classifiers performed well in isolation. Predicting the stroke status in patients would help the physicians determine the prognosis and assist them in providing the targeted therapy in a limited time. During this study, an ensemble model was built by considering the base, bagging, and boosting classifiers: Support Vector Machine, Naïve Bayes, Decision Tree, Logistic Regression, Artificial Neural Network, Random Forest, XGBoost, LightGBM, and CatBoost. The dataset consists of 5110 patient details, along with 12 attributes that were analyzed in this research. The final ensemble model was developed by carrying out the methodology in two phases. During the first and second phases, the classifiers mentioned above were trained without hyper-parameter tuning and with hyperparameter tuning and tested against the fundamental evaluation matrices. During each phase, the classifier that produces the highest classification accuracy is discovered from the base, bagging, and boosting categories. From the results obtained, the final ensemble model was constructed using the Max Voting approach, which yielded an accuracy of 95.76%.

Birth weight is a crucial measure of pregnancy outcome, and it indicates a neonate's chances of longevity, growth, long-lived health, and mental development. In epidemiological studies, it is commonly regarded as the causative pathway to these health consequences. Scanning the fetus's weight before birth is associated with difficulties and is typically calculated using empirical calculations based on doctors' clinical experience. In this research, the authors proposed an approach to predict the neonate weight class with minimum errors and costs during six months of pregnancy, utilising existing pregnant women's data. Authors recognised 16 attributes after an extreme literature survey and expert opinions. The haemoglobin level of the mother, mother's pre-pregnancy weight and height, pregnant woman's weight each month until 6 are some attributes identified. For the Machine learning (ML) models to work, a dataset of 1000 pregnant women was obtained and preprocessed. Then dataset was prepared to apply ML models. This research uses eight different ML algorithms to predict the neonate weight class at six months of pregnancy. Also, this research provides an approach to compare different eight ML models to find an effective algorithm for the neonate weight prediction. The Linear Discriminant Analysis (LDA) model has higher performance than other models. Accuracy, precision, recall and F1-score for LDA are 81.94 %, 83.05%, 94.23% and 88.28 % respectively.

Customer profitability is one of the most critical problems faced by businesses today. Keeping an existing customer is more valuable than gaining a new subscriber in the telecommunication industry. As a result, anticipating customer attrition behavior in advance is challenging. This behavior has prompted most researchers to establish a model for categorizing clients based on their profitability levels in various businesses. This study was carried out with the assistance of a local telecommunication service provider. Approximately 10,000 pre-paid subscriber details with 12 attributes were acquired. Furthermore, the classification technique was used to reduce the dimensionality between features and classify the high profitable customers, low profitable customers, and average profitable customers. The data was then fed into various supervised learning algorithms to choose the optimum algorithm by considering certain evaluation metrics for developing the final prediction model. The proposed approach revealed that the SVM outperformed all other techniques with greater accuracy of 80.00%.

Hundreds of lines of code make up large software systems. During maintenance and reuse, reading through the full source code will be quite tough. Many programmers stick to reading and navigating code than they develop it. Most experienced developers reuse existing code functionality. The cost of searching for and integrating reusable code determines the success of source code reuse. One method for identifying source code functionality names is to use a topic model. While numerous studies have focused on using topic models to analyze software systems, none have retrieved topics from actual source code syntax. By considering the optimal number of conditional statements, expressions, arrays, Boolean operators, and other unique programming keywords utilized is critical for any source code functionality name prediction. The dataset for the research model was collected from a variety of software developers' source code throughout the world. In software development, there are many different types of source code repositories, including source code algorithms. These repositories will be unstructured and unusable, requiring a technique to transform them into something useable. The source code data will be decomposed into smaller bits of programming language phrases using the Java parser library. Boosting algorithms are still more advantageous under the regime of limited training data and low knowledge for parameter modification. This study looked at supervised learning boosting algorithms such as XGBoost, CatBoost, LightGBM, and AdaBoost. It was discovered that the aforementioned algorithms performed equally, despite the fact that CatBoost surpassed all other methods based on accuracy with a value of 98%.

Researchers like to collaborate with other researchers to share their knowledge, experience, and resources with each other. By selecting appropriate collaborators researchers can obtain accurate output and they can publish more papers with high quality. They can get higher recognition within their research community. However, selecting appropriate collaborators is a challenging task. Thus, researchers proposed collaborator recommendation systems (CRS) to address that challenge. A network-based recommendation system is one type of CRS. Constructing an effective network is very important in the CRS. Existing network constructing approaches used co-author lists or publications to generate the network. However, the method is not suitable for some researchers like junior researchers and undergraduate students due to a lack of co-authors and publications. In this paper, we propose Google Scholar-based researchers network constructing approach. Common coauthors, the similarity of the area of interests, citation rate, and several co-authored publications between two researchers are extracted from GS to construct the network. The empirical work of our prototype system shows the efficacy of the technique presented.

Maintaining a customer base at a feasible rate is considered important in most business organizations since customers are the precious asset of the business sector. It is a vital task to retain the customers at a steady level in any business enterprise for the overall stability of its business activities. Sufficient pieces of evidence in this connection have been gathered to prove that the telecommunication industry is the most affected field of business by the tendency of the customers to shift towards alternative service providers. Therefore, a distinctive effort has been made to design this specific forecasting method is carried out utilizing a combination of a properly approachable method aiming at clarifying the probability of the above-mentioned tendency of the clients seeking an alternative service provider in the industry. In this attempt, a data set that included 10, 000 postpaid consumer particulars including 20 attributes were taken for this research for a thorough analysis of this aggravating issue in the telecommunication industry. In the end, a satisfying outcome was witnessed and certain clarification was made out of the 10,000 subscribers 4888 showed positive attitudes and 5112 indicated negative to the churning behavior. Besides, this specific data set was subjected to complete verification in comparison to certain supervised machine learning algorithms such as Decision tree, Logistic Regression, Support Vector Machine (SVM), and Artificial Neural Networks (ANN). Along with this, ensemble techniques such as Random Forest, Extreme Gradient Boosting (XGBoost), and Adaptive Boosting (AdaBoost) also have been considered. Subsequently, an assurance was made that XGBoost possessed the ability to bring out the maximum and précised accuracy of 82.90%. Eventually, a hyperparameter tuning had been performed with XGBoost. As a result, an assurance was acquired that XGBoost showed an upsurge in the previously obtained accuracy from 82.90% to 83.13%.

With the sudden upsurge of the Corona Virus (COVID -19) epidemic, higher education institutions around the globe have been engaging in untiring efforts to provide continuous learning opportunities for the students and to safeguard their welfare. This research is done to investigate the true experiences of university students in Sri Lanka using e-learning facilities throughout this period of the epidemic situation and forecast the adaption of e-learning in the universities after this epidemic situation. For this purpose, an online survey was designed and a formal questionnaire was circulated. In the end, after a thorough survey, a satisfactory result was obtained. Out of the 909 responses, 189 were not interested in involving in the e-learning process in the future whereas 460 were identified as satisfied with the adaption of e-learning. In addition, the rest of the 260 was satisfied with both the cases. Furthermore, the dataset fed into selected supervised machine learning techniques like Support Vector Machine (SVM) and Artificial Neural Network (ANN) to develop the prediction model on forecasting the likelihood of students with the adaption of the method of e-learning in the days to come. As per the results obtained, ANN outperformed against SVM and resulted in an accuracy of 76.19%. Immense inconveniences were experienced in the course of the students’ academic activities, because of the sudden upsurge of the pandemic. They are namely inaccessibility of technological devices, poor connectivity of the internet, and unsuitable surroundings for learning. This attempt also pinpoints the characteristics of the prevailing technological procedures in the making of distance learning theory. It was concluded that additional funding and feasible plans would be necessitated for the enhancement of an effective teaching-learning process towards the aforesaid method of the educational framework all over the country.

Diabetes Mellitus is commonly found in human beings around the world and this is one of the serious diseases which causes boundless suffering among patients. There are numerous reasons for the prevalence of this disease. It would be better to consider the predictions carried out earlier in this respect, since diabetes is a non - communicable disease and makes a great impact on the health condition of people nowadays. This is the reason why the existing medicinal practices in most hospitals are collecting patients' life history or the record of the disease. This is done for diagnosing diabetes using various medical tests followed by proper treatment for the disease. Machine learning provides an immense contribution to the sector of healthcare. For this research, Pima Indians Diabetes Dataset, obtained from the University of California, Irvine (UCI) machine learning source that included 768 patients' details along with nine attributes had been chosen for a comprehensive investigation of this grave and widespread problem in the health sector. Eventually, an adequate perfect outcome could be achieved and some effective and transparent conclusions were made. Among 768 diabetics, 500 were recognized as positive for the disease while 268 were recognized as negative. Besides, the recorded facts were put into particular supervised machine learning techniques such as Support Vector Machine (SVM), Naïve Bayes (NB), Decision Tree (DT), Artificial Neural Networks (ANN), Linear Discriminant Analysis (LDA), Logistic Regression (LR) and k-nearest neighbors (k-NN). Along with this, bagging and boosting techniques like Random Forest (RF), Extreme Gradient Boosting (XGBoost), LightGBM, and CatBoost too were taken into consideration. In addition, by considering classifiers with the highest accuracies, the final ensemble model was developed with the adaption of SVM, CatBoost and RF to predict the diabetes mellitus. Thus, the model resulted in an accuracy of 86.15%.