Browsing by Author "Ayman, Umme"

Now showing 1 - 3 of 3

A Machine Learning Based Approach to Classify Tense from English Text
(Scopus, 2024-12-19) Ayman, Umme; Islam, Md. Shafiqul; Rahat, Md. Azmain Mahtab; Raza, Dewan Mamun; Chakraborty, Narayan Ranjan; Bijoy, Md. Hasan Imam
This paper investigates the classification of tense in English text using machine learning algorithms. Support Vector Machine (SVM), Random Forest (RF), Multinomial Naive Bayes (MNB), Decision Tree (DT), XGBoost, and K-Nearest Neighbors (KNN) are the six classifiers used in the study. The dataset was collected from diverse sources including novels, books, blogs, articles, social media platforms, newspapers, websites and some of them self-made. The data underwent preprocessing steps such as cleaning, normalization, and feature extraction using TfidfVectorizer. Among the other algorithms, SVM achieved the highest accuracy at 97.17%. Classifier performance was assessed with metrics such as F1-score, recall, accuracy, and precision. To evaluate performance, ROC curves, and confusion matrices were also examined. The study underlines the necessity for focused approaches and draws attention to the significant gaps in the field of natural language processing (NLP) regarding tense classification studies. By leveraging machine learning, this research aims to enhance the accuracy and contextual appropriateness of tense classification, thereby improving cross-cultural communication and understanding in machine translation systems. This research contributes to NLP by offering a robust approach to tense classification and demonstrates the potential of SVM in achieving high accuracy for this task. Future work will focus on addressing limitations such as short training data, overfitting and tense conversion.
A Study on Drug Addiction Prediction in Bangladeshi Universities Using Advanced Machine Learning Algorithms
(Scopus, 2024-12-19) Ray, Binota; Ayman, Umme; Akash, Md Atik Asif Khan; Khan, Nusrat; Chakraborty, Narayan Ranjan; Bijoy, Md. Hasan Imam
Infection by alcohol and drugs has become one of the bigger dangers posed to the youth of Bangladesh, and responsible action on the part of society is called for in order to save these tender minds. In a bid to solve this problem, a study was conducted on reducing drug abuse using machine learning concepts. The data were gathered from 307 students, wherein there are drug users and non-users aged between 17-35, with the final dataset containing 21 features. Hence, the strategy for drug abusers can be predicted with the probability of an individual's addiction to drugs by designing a machine learning strategy. It is based on the consultations and opinions obtained from medical professionals and drug addicts, together with literature reviews, that key risk factors of addiction are suggested in the present study. The collected data was pre-processed and fed for four machine learning algorithms: logistic regression, SVM, naïve Bayes, and XGBoost. The performances of each classifier, as evaluated by some of the notable metrics, turned out to be: 95.16% for SVM, 93.55% for logistic regression, 96.77% for naïve Bayes, and 98.39% for XGBoost. The present research informs about the prospects of machine learning for risk assessment in drug addiction among the youth of Bangladesh and contributes to more effective prevention strategies. The main goal of this research is to develop predictive models to identify a person's risk of drug addiction based on behavioral, social and health factors. It can be an early intervention, prevention effort and help to improve its condition.
A Study on Drug Addiction Prediction in Bangladeshi Universities Using Advanced Machine Learning Algorithms
(Scopus, 2024-09-24) Ray, Binota; Ayman, Umme; Akash;, Md Atik Asif Khan; Khan, Nusrat
Infection by alcohol and drugs has become one of the bigger dangers posed to the youth of Bangladesh, and responsible action on the part of society is called for in order to save these tender minds. In a bid to solve this problem, a study was conducted on reducing drug abuse using machine learning concepts. The data were gathered from 307 students, wherein there are drug users and non-users aged between 17-35, with the final dataset containing 21 features. Hence, the strategy for drug abusers can be predicted with the probability of an individual's addiction to drugs by designing a machine learning strategy. It is based on the consultations and opinions obtained from medical professionals and drug addicts, together with literature reviews, that key risk factors of addiction are suggested in the present study. The collected data was pre-processed and fed for four machine learning algorithms: logistic regression, SVM, naïve Bayes, and XGBoost. The performances of each classifier, as evaluated by some of the notable metrics, turned out to be: 95.16% for SVM, 93.55% for logistic regression, 96.77% for naïve Bayes, and 98.39% for XGBoost. The present research informs about the prospects of machine learning for risk assessment in drug addiction among the youth of Bangladesh and contributes to more effective prevention strategies. The main goal of this research is to develop predictive models to identify a person's risk of drug addiction based on behavioral, social and health factors. It can be an early intervention, prevention effort and help to improve its condition.