Journal of Computer Science and Its Application

Comparative Analysis of Clustering Algorithms on High Dimensionality Data

M. Abdulraheem, I.D. Oladipo, G.B. Balogun, M.O. Adeleke, D.S. Ricketts — Mon, 07 Aug 2023 00:00:00 +0000

Data mining is an emerging research area employed by many evolving computing technologies since it reduces dataset complexity by providing remarkable insight into the data. Additionally, it requires the ability to creatively envision the enormous and heterogeneous datasets and to extract meaningful knowledge from the plethora of data through the practical application of appropriate algorithms. For this reason, clustering algorithms are categorized as hierarchical, partitioning, and density-based and grid-based. The Partitioning Clustering technique divides the data objects into several groups known as partitions, and each division represents a cluster. A hierarchy or tree of clusters is created for the data objects using hierarchical clustering algorithms. The cluster is in areas with high densities by density-based algorithms, which aggregate their data objects based on a particular neighbourhood. The grid structure used by a grid- based algorithm is created as the data object space is divided into a finite number of cells. Moreover, clustering is a technique that is frequently used in data mining to examine the data; thus the authors were motivated to compare it with other approaches. A data mining analysis is useful for gaining an understanding of the distribution of data, observing the characteristics of clusters, and focusing on certain clusters for further analysis. This work focuses on determining the algorithm with better performance on high-dimensionality data between Expectation Maximization (EM) and Hierarchical Algorithms (HA) using cluster accuracy and evaluation time as parameters for comparison. In this study, cluster analysis was performed using WEKA 3.8.5. The result shows that the EM method runtime and accuracy perform better in clustering high-dimension data and performance improves as the number of clusters increases. However, in the HA method, running time and accuracy barely improved with the difference in the dataset. Therefore, it is observed that the HA method falls short in performance compared to the EM method.

A Comparison of LSTM and Some Machine Learning Regressors for the Prediction of Close Prices of the Nigerian Stock Market

A.U. Rufai, E.P. Fasina, E.O. Oduola — Mon, 07 Aug 2023 00:00:00 +0000

A lot of work has been done on stock market predictions in the literature. A salient fact is that stock market prediction is a dynamic phenomenon. This is due to the fact that once a prediction model is known, it becomes necessary to device a new model due to the dynamic nature of the market. Traditional predicting method is characterized by the use of fundamental and technical analysis. Fundamental analysis is based on economy indices of the market while the technical analysis is based on history and trends of the market. In this paper, we present the application of recurrent neural network for the prediction of the close price of stocks in the Nigerian stock market. We confirm previous results that states that deep learning algorithms offer a higher performance than traditional machine learning algorithms like support vector regressor, catboost regressor and multi-linear regressor. The experimental result obtained shows that RNN performed best with the least mean squared error of 0.036577 while multi-linear regressor has mean squared error of 0.049327, support vector regressor has mean squared error of 0.066206 and catboost regressor has mean squared error of 0.091510

Smartphone Usage among Computer Science Students in Higher Education during Covid-19 Lockdown

M.O. Edeh, S.G. Ugboaja, N.E. Ugwuja, J.S. Igwe, I.E. Daniel, N.E. Richard-Nnabu — Mon, 07 Aug 2023 00:00:00 +0000

Smartphone is fast emerging as an important pedagogical tool due to its features and capabilities. The study investigated Smartphone use among Computer Science students in higher education during the COVID-19 lockdown with a view to ascertain its effect on their academic activities. A total of 200 questionnaires were distributed online to both undergraduate and postgraduate students studying Computer Science at various higher institutions. The result shows that ownership of Smartphone was widespread among the participants and its use had statistically significant effects on the academic activities of the participants. Most of the participants admitted that their Smartphones enhanced their access to the internet, school updates, digital learning, communication, research and content creation during the lockdown. Similarly, student’s attitudes towards Smartphone use were also found to be pleasant, and it had statistically significant effects on their extent of usage for academic activities during the COVID-19. However, bad network, small size screen etc were found to inhibit Smartphone use for academic activities.

A smart-based dustbin for office waste management

Mon, 07 Aug 2023 00:00:00 +0000

Smart waste management systems have become increasingly popular in recent years, with a focus on developing more efficient and sustainable waste management solutions. In particular, office waste management is a critical issue that requires more effective solutions. This paper presents a novel approach for a smart dustbin system based on a Convolutional Neural Network (CNN) framework, with secured Bluetooth connectivity using Elliptic Curve Cryptography (ECC) in sensitive situations. The proposed system is designed to detect the level of waste in the dustbin, its gas purity, and recognize voice commands for automatic mobility. The CNN model is trained on a dataset of spectrograms of speech signals to enable the detection and recognition of voice commands. The gas purity level is detected using a gas sensor placed inside the dustbin, while an ultrasonic sensor is used to measure the waste level. The CNN model and sensor data are integrated with an Arduino board to send notifications to a mobile application via Bluetooth connectivity with ECC for secured data transmission. The proposed system was evaluated using a prototype implementation, and the results showed that the CNN model achieved high accuracy in speech recognition, while the waste level detection and gas purity detection were accurate and efficient. The use of ECC provided secured data transmission, which is crucial in protecting user privacy and preventing data tampering. The design was evaluated with a usability experiment based on four key performance metrics. The experimental process provided an average of 80% of participants' approval of the proposed system. Thus, the proposed smart dustbin system has the potential to improve waste management efficiency by providing timely notifications to waste collection services and reducing environmental pollution. Future research can explore the use of machine learning algorithms to adopt waste segregation in the smart dustbin system.

Comparative Evaluation of Image Segmentation Techniques for Flood Detection in High-Resolution Satellite Imagery

C. Agbo, A.D. Mohammed, J.K. Alhassan, S.A. Adepoju — Mon, 07 Aug 2023 00:00:00 +0000

Speedy reaction to natural disasters, such as floods, is critical to minimising loss of life and pain. Access to fast and reliable data is critical for rescue teams. Satellite photography provides a wealth of data that may be analysed to assist pinpoint disaster-affected areas. The use of segmentation to analyse satellite images is becoming increasingly important in environmental and climatic monitoring, particularly in detecting and controlling natural disasters. Image segmentation improves pattern recognition, which divides a single image into several homogeneous pieces. The efficiency of image segmentation techniques varies depending on the layout of objects, illumination, shadow, and other variables. However, there is no one-size-fits-all method for successfully segmenting all imagery; specific methods are more efficient than others. This report compares four different technologies. Commonly used image segmentation techniques: K-means clustering (K.C.), Color thresholding (C.T.), Region-based Active Contour (R.A.C.) and Edge-based Active Contour (E.A.C.) segmentation. These four techniques were used to detect, and segment flooded areas in high-resolution satellite imagery. The K.C. method had the best flood segmentation rate with a Jaccard Index of 0.8234, Dice of 0.9234, the precision of 0.9589, recall of 0.9078 and BFscore of 0.9327, which was higher than the other three segmentation technique and previous works.

Conceptual Framework for African-Centric Geolocation Data Generation, Mapping and Management

G.I. Okonkwo, I.I. Umeh, I.E. Ezeani — Mon, 07 Aug 2023 00:00:00 +0000

Geolocation data can be used to determine the position of any object on the planet earth with reasonable exactness. This position is determined by the Latitude and Longitude values of a point on the globe. Though most current smart devices now have in-built navigation systems and digital maps, the maps does not incorporate accurate Africa local content required and consequently, lack the desired detail and accuracy required to be of efficient use in Africa. Reviews of existing solutions were made and they proved not to be of optimal benefit to Africa. A framework to address this issue is presented. The objective is to have a localized Information System that will supersede some of the existing solutions to boost the African-Centric Geolocation Database and this will in turn contribute to the body of knowledge in the context of Location Finding and Navigation to provide the “missing local contents” over time.

TomatoDetect: A ConvNet-Powered Mobile Application for Detecting Tomato Leaf Diseases

Mon, 07 Aug 2023 00:00:00 +0000

Tomatoes are a mainstay in Nigerian cuisine, showing up in a variety of cuisines and offering several nutritional advantages like Vitamin C, potassium, and lycopene, which guard against cancer and heart disease. Nigeria now produces 10.8% of all fresh tomato production on the continent, placing it second overall. To avoid the enormous loss, extra care must be taken with the vulnerable crop tomato plant because of many infections that affect it. Farmers and other agricultural experts undergo laborious and time-consuming procedures when visually checking crops that they think to be impacted by various infections in the real world, which does not ensure proper recognition and classification of particular plant diseases. To identify healthy tomato leaves and nine tomato leaf infections, this study created a mobile application. This study created two pre-trained VGG-16 Convolutional Neural Networks (CNN or ConvNet) models using the Keras deep learning framework. With an accuracy of 96.51%, the model trained on the enhanced data surpassed the model trained without the augmented data. To accurately detect specific diseases and categorize healthy leaves in a real-world scenario in tomato leaves, this DL model was chosen and used in a built mobile application. The chosen VGG-16 pre-trained model was first transformed into a TensorFlowLite (TFLite) model applicable in an android mobile application before being deployed into a mobile application environment. The mechanism of gathering user data and transmitting it through the backend to be verified with the firebase database, which manages the application's storage and authentication, was designed using the kotlin programming language. With this mobile app in their hands, tomato farmers can detect disease outbreaks and spread in tomato leaves before they run out of control and endanger food security.