Journal of Information Technology and Computing

Ontology based Feature Selection and Weighting for Text classification using Machine Learning

Djelloul BOUCHIHA — 2023-06-27

Text classification consists in attributing text (document) to its corresponding class (category). It can be performed using an artificial intelligence technique called machine learning. However, before training the machine learning model that classifies texts, three main steps are also mandatory: (1) Preprocessing, which cleans the text; (2) Feature selection, which chooses the features that significantly represent the text; and (3) Feature weighting, which aims at numerically representing text through feature vector. In this paper, we propose two algorithms for feature selection and feature weighting. Unlike most existing works, our algorithms are sense-based since they use ontology to represent, not the syntax, but the sense of a text as a feature vector. Experiments show that our approach gives encouraging results compared to existing works. However, some additional suggested improvements can make these results more impressive.

A Lightweight Machine Learning-Based Email Spam Detection Model Using Word Frequency Pattern

Mohamed Aly Bouke — 2023-06-27

This Spam emails have become a severe challenge that irritates and consumes recipients' time. On the one hand, existing spam detection techniques have low detection rates and cannot tolerate high-dimensional data. Moreover, due to the machine learning algorithm's effectiveness in identifying mail as solicited or unsolicited, their approaches have become common in spam detection systems. This paper proposes a lightweight machine learning-based spam detection model based on Random Forest (RF) algorithm. According to the empirical results, the proposed model achieved a 97% accuracy on the spambase dataset. Furthermore, the performance of the proposed model was evaluated using standard classification metrics such as Fscore, Recall, Precision, and Accuracy. The comparison of Our model with state-of-the-art works investigated in this paper showed the model performs better, with an improvement of 6% for all metrics.

Data Mining and Business Intelligence for Better Healthcare Decision

Sk Tanvir Ahmed — 2023-06-27

Business intelligence is a subject of business information whose purpose is to make strategies that include new technologies, applications, and practices to collect the targeted information. Data mining is one of the most motivating areas of research and it is become gaining popularity in healthcare organization day by day. Data mining is based on several techniques such as classification, clustering, association, and regression in the health domain. While data mining has several advantages but also has disadvantages. This research-based finding helps any healthcare organization to make the decision that helps the organization to become more popular and demandable.

Diagnosing Chronic Kidney disease using Artificial Neural Network (ANN)

Ala rashid — 2023-06-27

The prevalence of chronic kidney disease (CKD), brought on by environmental pollution and a lack
of safeguards for people's health, is rising globally. A slow and steady decrease in kidney function
over many years is chronic kidney disease (CKD). A person may eventually get renal failure. Using
artificial neural networks in concert with the machine learning techniques (ANN), Keras, and Google
Colab Notebook for serial model construction, this study intends to propose a potent method for
identifying chronic kidney disease.
This study looked into ANN's accuracy, sensitivity, and specificity in the diagnosis of CKD. Based
on the dataset's purpose, categorization of technology's effectiveness. In order to decrease the feature
dimension and increase classification system accuracy, an algorithm model including ANN has been
developed.
Results indicate that ANN architecture, which was used, achieved the best accuracy (98.56%),
whereas other methods, such as SVM, Random-forest, and K-Nearest Neighbor (KNN), delivered
accuracy levels that were lower than those of ANN.

Quality challenges in Deep Learning Data Collection in perspective of Artificial Intelligence

Gowri Vidhya — 2023-06-27

With reinforcement learning powered by big data and computer infrastructure, data-centric AI is driving a fundamental shift in the way software is developed. To treat data as a first-class citizen on par with code, software engineering must be rethought in this situation. One surprise finding is how much time is spent on data preparation throughout the machine learning process. Even the most powerful machine learning algorithms will struggle to perform adequately in the absence of high-quality data. Advanced technologies that are data-centric are being used more frequently as a result. Unfortunately, a lot of real-world datasets are small, unclean, biased, and occasionally even tainted. In this study, we focus on the scientific community for data collecting and data quality for deep learning applications. Data collection is essential since modern algorithms for deep learning rely mostly on large-scale data collecting than classification techniques. To enhance data quality, we investigate data validation, cleaning, and integration techniques. Even if the data cannot be completely cleaned, robust model training strategies enable us to work with imperfect data during training the model. Furthermore, despite the fact that that these issues have gotten less attention in conventional data management studies, bias and fairness are significant themes in modern application of machine learning. In order to prevent injustice, we investigate controls for fairness and strategies for doing so before, during, and after model training. We believe the information management community is in a good position to address these problems.