Social media platforms such as Facebook, LinkedIn, Twitter, among others have been used as tools for staging protests, opinion polls, campaign strategy, medium of agitation and a place of interest expression especially during elections. In this work, a Natural Language Processing framework is designed to understand Nigeria 2023 presidential election based on public opinion using Twitter dataset. Two million tweets with 18 features were collected from Twitter containing public and personal tweets of the three top contestants – Atiku Abubakar, Peter Obi and Bola Tinubu – in the forthcoming 2023 Presidential election. Sentiment analysis was performed on the preprocessed dataset using three machine learning models namely: Long Short-Term Memory (LSTM) Recurrent Neural Network, Bidirectional Encoder Representations from Transformers (BERT) and Linear Support Vector Classifier (LSVC) models. This study spanned ten weeks starting from the candidates’ declaration of intent to run for Presidency. The sentiment models gave an accuracy, precision, recall, AUC and f-measure of 88%, 82.7%, 87.2%, 87.6% and 82.9% respectively for LSTM; 94%, 88.5%, 92.5%, 94.7% and 91.7% respectively for BERT and 73%, 81.4%, 76.4%, 81.2% and 79.2% respectively for LSVC. Result also showed that Peter Obi has the highest total impressions the highest positive sentiments, Tinubu has the highest network of active friends while Atiku has the highest number of followers. Sentiment analysis and other Natural Language Understanding tasks can aid in the understanding of the social media space in terms of public opinion mining. We conclude that opinion mining from Twitter can form a general basis for generating insights for election as well as modeling election outcomes.

Diabetes mellitus is a prevalent chronic illness with severe complications that demand timely diagnosis. This study introduces an attention-enhanced Deep Belief Network (DBN) for early diabetes risk prediction, designed to address challenges associated with highly imbalanced datasets. Using a dataset from Sylhet Diabetes Hospital, which includes symptom and demographic information from patients, we applied an ensemble feature selection approach to identify critical predictors. To address the class imbalance, Generative Adversarial Networks (GANs) were used to generate synthetic data, ensuring the model’s robustness in identifying underrepresented cases. Additionally, a hybrid loss function combining cross-entropy and focal loss was implemented to improve classification, especially for hard-to-detect instances. Our results show that the attention-based DBN model, augmented with synthetic data from GANs and optimized with a hybrid loss function, achieves an AUC of 1.00, F1-score of 0.97, precision of 0.98, and recall of 0.95, outperforming several baseline models. This research offers a novel and effective approach for early diabetes detection, demonstrating potential for use as a clinical tool in preventive healthcare settings.

Introduction: Prostate cancer (PCa) is one of the deadliest and most common causes of malignancy and death in men worldwide, with a higher prevalence and mortality in developing countries specifically. Factors such as age, family history, race and certain genetic mutations are some of the factors contributing to the occurrence of PCa in men. Recent advances in technology and algorithms gave rise to the computer-aided diagnosis (CAD) of PCa. With the availability of medical image datasets and emerging trends in state-of-the-art machine and deep learning techniques, there has been a growth in recent related publications. Materials and Methods: In this study, we present a systematic review of PCa diagnosis with medical images using machine learning and deep learning techniques. We conducted a thorough review of the relevant studies indexed in four databases (IEEE, PubMed, Springer and ScienceDirect) using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. With well-defined search terms, a total of 608 articles were identified, and 77 met the final inclusion criteria. The key elements in the included papers are presented and conclusions are drawn from them. Results: The findings show that the United States has the most research in PCa diagnosis with machine learning, Magnetic Resonance Images are the most used datasets and transfer learning is the most used method of diagnosing PCa in recent times. In addition, some available PCa datasets and some key considerations for the choice of loss function in the deep learning models are presented. The limitations and lessons learnt are discussed, and some key recommendations are made. Conclusion: The discoveries and the conclusions of this work are organized so as to enable researchers in the same domain to use this work and make crucial implementation decisions.

Twitter (X) has emerged as a vital platform for understanding public opinion and forecasting election outcomes through real-time sentiment analysis. This study systematically reviews machine learning (ML) and deep learning (DL) techniques applied to Twitter data for election prediction. Following PRISMA guidelines, 275 studies were screened, and 76 met the inclusion criteria. The review identifies dominant methods, performance trends, and research gaps in this field. Results show that pre-election data are most frequently analyzed, with the United States, India, and Indonesia leading the research landscape. Term Frequency–Inverse Document Frequency (TF–IDF), often paired with n-gram models, remains the most widely adopted feature extraction technique. Naïve Bayes and Support Vector Machines consistently deliver strong predictive performance, while newer DL models such as BERT and Bi-LSTM demonstrate improved contextual understanding. Approximately 78% of reviewed studies correctly predicted real election outcomes. This review consolidates methodological evidence, highlights emerging transformer-based innovations, and outlines directions for multilingual, cross-platform, and explainable AI approaches in electoral forecasting.

Large-language-model reasoning is commonly represented as a serial transcript or as a tree of textual alternatives, even when many distinct trajectories reach the same executable state. We present Interference Search, a state-based search procedure that expands a bounded frontier synchronously, executes candidate actions in an environment, merges future-equivalent states, and optionally applies a learned viability judge before advancing the next frontier. We formalize the exact component of the method as search on a quotient state space and prove that canonical state merging preserves bounded-depth goal reachability whenever the equivalence relation preserves goal membership and future transitions. We further define the path-to-state compression factor and show that exponential representational savings are possible when path growth is exponential while the number of reachable equivalence classes is polynomial. Learned pruning is treated separately as an explicitly lossy approximation, yielding a simple survival bound in terms of the judge's false-negative rate. Empirically, six-number Countdown search produced 831,176 operation paths but 13,229 distinct states (62.8 paths per state). On 30 hard four-number Countdown instances, with the same judge and a budget of 200 judged positions, the merged synchronous frontier solved 30/30 compared with 21/30 for a single-line baseline. In a 100-problem six-number ablation at 100 expansions, the merged frontier solved 77\%, compared with 50\% without merging and 40\% for the single-line search. A preliminary MBPP experiment produced 9/30 solutions forInterference Search, 8/30 for independent sampling, and 7/30 for two linear refinement strategies; this difference is not interpreted as statistically established superiority. The results support a narrower claim: future-preserving state merging can remove substantial redundant reasoning while retaining reachability, whereas learned pruning introduces a measurable compression--completeness trade-off.

Language-model reasoning systems can spend substantial computation on trajectories that converge to the same environment state. This paper develops a verifier-grounded account of Interference Search as quotient-state search with separate operators for exact merging, certified exclusion, learned inhibition, and execution scheduling. We give sufficient conditions for goal-preserving state equivalence, prove witness preservation and sound backward refutation, derive a pathwise bound on learned-pruning loss, and establish work--latency trade-offs for batched verification. The empirical study synthesizes supplied experiment reports rather than independently rerunning their remote receipts. A 1,000-instance Countdown protocol reports that conventional best-first search exceeds the full width-16 frontier at tight state-work budgets, defeating a universal dominance claim. In a native Qwen3-4B online evaluation on 200 five-number instances, exact merging improves solve rate from 16.5% to 27.0%, with 22 paired gains and one loss. Learned inhibition yields a smaller 4.5-point improvement over the same merged checkpoint with coupling disabled. A separate 64-task MBPP runtime replication preserves 47 visible-test solves while reducing mean completion latency from 2.265 to 2.064 seconds; its reported paired saving interval is 0.136--0.276 seconds. Native midtraining and latent-slot experiments expose capability regression and incomplete hypothesis decoding. These results support exact sharing and workload-specific scheduling benefits, while leaving broad reasoning superiority and autonomous latent hypothesis tracking for future exploration.