ArticleResearch Article
MindBridge: A Transformer-Based Conversational Agent for Real-Time Depression and Anxiety Detection and Support in University StudentsJoseph Oluwaseyi Adewuyi, Onwualu Nwamaka Victoria, Ogbonna Jordan Eme, Madagwa Success
August 28, 2026•DOI: 10.52589/BJCNIT-9LRO2SCJ•Volume 9 Issue 2•313 views
Mental health challenges are becoming a major concern among university students who still have limited access to appropriate and affordable support due to barriers such as stigma, cost, and time limitations. Most of the digital tools available, such as Wysa and Woebot, are based on rigid rule-based logic and are incapable of understanding the contextual nuances of student-expressed emotional distress.
For this system, a system-oriented, agile research methodology was used. Three public datasets the Mental Health Text Classification Dataset, the Reddit Depression Dataset, and the EmpatheticDialogues Dataset, were combined and balanced to create a training corpus of 24,000 labeled entries equally distributed across three emotional categories: depressive, anxious, and neutral. The Mental-BERT transformer model was trained on this corpus in five epochs with the help of Google Colab's GPU infrastructure. This classification model was then incorporated into a Flask web backend, together with the Groq API for conversational response generation, a PostgreSQL relational database for secure data persistence, and a web-based frontend consisting of a responsive chat interface with user authentication.
After evaluating the fine-tuned Mental-BERT model on a test set of 3,600 samples that were not used during training, the overall classification accuracy was 96%. Also, the macro-averaged precision, recall, and F1-score were equally high at 96% across the three classes. We carried out a user study with 58 university students who gave a mean System Usability Scale score of 76.55 for MindBridge. That score belongs to the good usability category and is also above the industry average threshold of 68. More than 70% of participants graded the system as Good or Excellent. Empathy and conversational quality metrics showed that the respondents considered MindBridge a source of empathy and assistance, while the mean figure for perceived usefulness was 4.03 out of 5. All 16 automated system tests were successfully passed.
Results show that using a domain-specific pre-trained transformer model alongside prompt-engineered large language model response generation and a hybrid safety architecture can lead to the development of a very accurate, usable, and empathetic mental health support system for university students.
Read moreMindBridgeTransformerConversational AgentDepressionAnxietyDetectionMental-BERT ModelMental HealthEmotional Classification
ArticleResearch Article
Comparative Study of Linear Discriminant Analysis (LDA), Quadratic Discriminant Analysis (QDA) and Support Vector Machine (SVM) In Dataset.Okoh Jophet Ewere, Owoyi Mildred Chiyeaka, Okoh Jophet Ewere
March 10, 2025•DOI: 10.52589/AJSTE-UAZPPMER•Volume 5 Issue 1•1,377 views
Classification techniques is an important factor in data analysis. Over the years, different classification method have been proposed for classification of dataset. In this paper, we compared three classifiers (LDA, QDA and SVM) in three imbalanced datasets (Iris, Pima and Glass data) and misclassification rate of the three classification method were compared. The experiments concentrated on analyzing the average misclassification rate among classifiers across the three dataset studied using the misforest imputation method to balance the dataset respectively. The results reveal that for the glass dataset, the QDA classifies the dataset better than the two other classification method studied, while for the iris and glass datasets, the LDA outperformed the other two classifiers studied. The conclusion in this study is that LDA have the least average misclassification error, followed by the QDA and then the SVM with an average misclassification rate of 0.2863.
Read moreimbalanced data; misclassification rate; Average misclassification; classification.
ArticleResearch Article
Spatial Capability Classification and Land Evaluation Studies for Rice Production on Selected Wetland Soils in Ekiti State, Nigeria.Ajayi Samuel Ojo, Fasina Sunday Abayomi, Babalola Temitope Seun, Ilori Augustus Oludotun, Fakorede Bolaji Martins
January 27, 2025•DOI: 10.52589/AJAFS-Q1W0H87B•Volume 8 Issue 1•1,707 views
The evaluation of wetland soils is essential for understanding their suitability for various agricultural practices, particularly in regions where soil properties significantly influence land management decisions. This study focuses on the spatial capability classification and land evaluation of selected wetland soils in Ekiti State, Nigeria, with a particular emphasis on rice production. Wetland soils in this region are characterized by distinct physical and chemical properties that affect their agricultural potential. The use of both parametric and non-parametric approaches in land suitability assessment allowed for a comprehensive analysis of the land's potential, factoring in both inherent soil qualities and the potential for improvement through human intervention. The research utilized a combination of Land Capability Classification (LCC) and Land Suitability Classification (LSC) methods to assess soil characteristics such as texture, organic matter content, drainage capacity, and nutrient availability. These evaluations were further supported by the Geographic Information System (GIS) and remote sensing technologies, which provided spatial data and generated suitability maps. The results indicate a range of soil suitability classes, from highly suitable (Class I) to unsuitable (Class VIII) for rice cultivation. Although, the drainage condition which is essential for effective water management indicated moderate suitability (S2) in all the locations. Notably, sites such as Ise, Isan, and Ikole demonstrated highly suitable conditions (S1) for rice production, characterized by optimal soil textures. In contrast, Ifaki exhibited marginal suitability (S3) due to lower pH levels and cation exchange capacity (CEC), highlighting potential challenges in nutrient retention that could adversely impact crop yield. The results of this study offer valuable insights for land use planning and agricultural development, contributing to sustainable land management practices and mitigating erosion risks, especially in regions identified as having moderate to severe erosion hazards and also enhanced food security in the region. Additionally, it emphasizes the need for continuous soil assessments to monitor changes in soil properties and adapt agricultural strategies accordingly.
Read moreSpatialevaluationwetlandclassification.
ArticleReview Article
A Review on the Effect of Imbalanced Dataset on Linear Discriminant Analysis.Owoyi M. C., Okwonu F. Z.
November 25, 2024•DOI: 10.52589/AJMSS-ZOZBNYPR•Volume 7 Issue 4•1,308 views
Imbalanced data are often delegate issues in data set as it has the power to affect the result and the performance of the classification algorithm. In such problems, if not handled well with a good sampling techques could lead to biased result, overfitting as well as high rate of misclassification thereby favouring just one class among the two classes. Usually, when assigning a sampling techniques, it is necessary to look at the nature of the dataset being studied. It is of a truth that the LDA classifier looking for an efficient performance when presented with an imbalanced instances are not suitable to deal with imbalanced learning tasks, since they tend to classify all the data into the majority class, which is usually the less important class. This work explains the different approaches which have been employed by different researchers to resolve the issue of imbalanced data in LDA and the effect of the result obtained both positively and negatively. It should be noted that this single article cannot completely review all the works or researches done on the topic, hence we hope that the references which was dually cited will be of help to the major theoretical issues.
Read moreImbalanced data; Oversampling; Undersampling; Classification; Metric evaluation.
ArticleResearch Article
On the Comparative Study of the Supervised Machine Learning Models.Asogwa O. C., Eze N. M., Christopher C. C., Aruah K. E., Chimezie I. E.
November 14, 2024•DOI: 10.52589/AJMSS-ILF4K7JB•Volume 7 Issue 4•1,241 views
This research empirically compared the performance of three supervised machine learning models which are Multinomial Logistic Regression (MLR), Multilayer back propagated Neural Networks (MNN) and Multinomial Decision Trees (MDT) based on a tested data set for the opinion of Nigerian citizens during the period of naira redesign policy using classification matrix criterion. About 600 copies of questionnaires on the opinion of Nigeria citizens on their individual welfare at the period of naira redesign. The result showed that, ANN outperformed other models with 94.4% correct classification rates, followed by the MLR with correctly classification rates of 93.5% and lastly by MDT with correct classification rates of 90.0%.
Read moreSupervised machine learningAlgorithmNaira redesignClassification ratesModels.
ArticleMeta Analysis
MRI-Based Brain Tumour Classification Using Convolutional Neural Networks: A Systematic Review and Meta-AnalysisErnest E. Onuiri, Adeyemi John, Kelechi C. Umeaka
October 9, 2024•DOI: 10.52589/BJCNIT-LOYYI2RS•Volume 7 Issue 4•1,714 views
This study systematically reviewed advancements in brain tumor classification using convolutional neural networks (CNNs) with MRI data, aiming to assess the effectiveness and potential for future enhancements. An analysis of 37 studies demonstrated the extensive use of CNN architectures and pre-processing techniques, achieving high classification accuracy rates. However, challenges such as class imbalances and model interpretability were identified. To address these issues, the study recommends further exploration of advanced deep learning techniques, ensemble methods, and the inclusion of more diverse datasets. A maximum accuracy of 98.80% was reported on a dataset comprising 154 MRI brain images, demonstrating the effectiveness of CNNs in brain cancer detection. Additionally, a five-year meta-analysis (2018-2022) on MRI scan cases across different demographic groups revealed important patterns in healthcare resource allocation. This research not only provides a comprehensive evaluation of CNN usage in MRI-based brain tumor classification but also outlines directions for future research to enhance diagnostic capabilities.
Read moreConvolutional neural networksdeep learningMRI databrain tumour classificationaccuracysystematic reviewhealthcare trends.
Article
Characterization, Classification and Suitability Evaluation of Soils in Ukum, Benue State, Nigeria for Maize and Yam Production.Apelebiri S., Ukabiala M. E., Asadu C. L. A., Ezeaku P. I., Ajoagu G. M.
July 19, 2024•DOI: 10.52589/AJAFS-JKFC8H88•Volume 7 Issue 3•2,105 views
Soil characterization, classification and evaluation provide useful information for the understanding of soil, its potentials and limitations for crop production. The study was to determine the morphological, physical and chemical properties of the soils; classify the soils; and evaluate their suitability for maize, cassava and yam production. Three land utilization types (maize, cassava and yam) with a control plot each were used for the study. One soil profile pit was dug in each land utilization type and control. The pedons were described in-situ following the procedures in FAO, s guidelines for soil profile description. A total of 23 samples were collected from the diagnostic soil horizons. The soils were classified using USDA Soil taxonomy and correlated with FAO/UNESCO World Resource Base. The suitability of the soils for maize, cassava and yam production was assessed using the non parametric and parametric methods of land evaluation. Morphologically, the soils were well drained and very deep with predominantly brownish gray (5YR5/1) and dark gray (2.5Y 5/2). The soil structures varied from weak fine granular structure to weak/moderately medium angular and sub-angular blocky structures. Soil texture varied from sand to sandy loam. The soil bulk density and total porosity varied from 1.42 to 1.72 g cm-1 and 36.98 to 46.42 %, respectively. The pH of the soils varied from slightly acidic to moderately acidic. Organic matter and total nitrogen in the soils were rated very low to low. The soils were classified at various great group levels of ultisols. The soils were moderately suitable.
Read moreOrganic matterSoil pHCharacterizationClassificationPedon.
Article
Characterization and Classification of Soils Along Three Toposequences in Njikoka Area, Anambra State, Nigeria.Ajoagu G. M., Ukabiala M. E., Asadu C. L. A., Umeugokwe P. C.
July 18, 2024•DOI: 10.52589/AJAFS-F6PREDRA•Volume 7 Issue 3•2,065 views
The study aimed to characterize and classify soils in the Njikoka Area of Anambra State, Nigeria. Three toposequences (TA, TB, and TC) from three different communities: Abagana, Nimo and Enugu-ukwu were selected and divided into upper (US), middle (MS), and lower (LS) slopes using a Digital Elevation Model (DEM). Nine pedons were described following FAO guidelines, and 64 samples were collected for analysis. The soils were well-drained and very deep, with predominant dark reddish brown and reddish brown horizons. Soil structure varied from weak fine granular to moderately medium angular and sub-angular blocky structures, while texture varied due to variations in sand, silt, and clay fractions. Soil acidity ranged from extremely acidic to slightly acidic. Organic matter, total nitrogen, available phosphorus, and exchangeable bases were generally low to very low. Potassium content ranged from very low to moderate, while exchangeable acidity, base saturation, and cation exchange capacity (CEC) varied from very low to high. The trend in chemical fertility was MS > US > LS for TA, US > MS > LS for TB, and LS > MS > US for TC soils. Based on the USDA soil classification, TA soils were identified as Typic Hapludults (US), Typic Kandiudults (MS), and Aquic Arenic Hapludults (LS). TB soils were classified as Grossarenic Kandiudualf (US) and Typic Kandiudults (MS and LS), while TC soils were categorized as Vertic Hapludults (US), Grossarenic Hapludults (MS), and Grossarenic Kandiudults (LS).
Read moreToposequenceCharacterizationClassificationTopographic unitsNjikoka.
Article
Fake News Detection System Using Logistic Regression, Decision Tree and Random Forest.Oni Oluwabunmi Ayankemi, Idowu Oluwaferanmi Ruth, Bassir Abdullai Abiye
May 17, 2024•DOI: 10.52589/BJCNIT-IOYRPY7G•Volume 7 Issue 1•3,000 views
The purpose of this study is to design a fake news detection system with these three machine learning models, namely: Decision Tree, Random Forest, and Logistic Regression. These three different models were analysed to determine the most efficient model for accurately detecting fake news. The result obtained showcased Logistic Regression with an accuracy of 98.80%, Decision Tree with an accuracy of 99.64% and Random Forest with an accuracy of 99.23%. It is evident as deduction from the comparative analysis that our best model came out to be Decision Tree with an accuracy of 99.64%.
Read moreInternetSocial mediaFake newsClassificationMachine learningData setComputational technique.
Article
Comparative Analysis of Weather Prediction Using Classification Algorithm: Random Forest Classifier, Decision Tree Classifier and Extra Tree Classifier.Oni Oluwabunmi Ayankemi, Iskilu Zainab Adesola, Lawrence Adeolu
May 17, 2024•DOI: 10.52589/AJMSS-F6H03BNE•Volume 7 Issue 2•1,694 views
Comparison of machine learning models is carried out in order to determine which models are best to deploy as a system. However, for the purpose of our research we carried out a comparative analysis on Random Forest classifier, Decision Tree classifier and Extra Tree classifier for weather prediction systems as we focused on seeking the classifier with the highest performance metrics. Based on the metrics, accuracy score, the best model for the system was determined. We carried out training, testing and validation of the three different models on the same dataset from the Kaggle dataset. We were able to implement Random Forest Classifier, Decision Tree Classifier and Extra Tree Classifier from Scikit-Learn to make weather prediction and using matplotlib to visualize the accuracy score of the implemented models. The Random Forest Classifier was chosen as the best able to achieve the highest at 66% accuracy.
Read moreWeatherPredictionClassificationDecision treesRandom forestLogistic regressionSupport vector machine.