International Journal of Internet of Things and Smart Computing Environment

An Intelligent Web-Based Medical Management and Prescription Recommendation System Using Flask, MySQL, and Machine Learning

Patange S. P, Bhadugale O. U, Bagwan S. F, Naladwade A. S — Tue, 07 Jul 2026 00:00:00 +0000

The healthcare sector continues to struggle with coordinating appointment scheduling, prescription management, patient feedback, and emergency-service delivery across disconnected, often manual, workflows. Addressing these issues, this study reports on the formulation, construction, and testing of an intelligent medical management system, a Flask-based web application supported by a MySQL database that brings together the four main users (patient, doctor, druggist, and administrator) in one single, secure, and role-based platform. The system is equipped with a supervised machine-learning component that suggests medications and personalised diet plans based on patient-reported symptoms as well as historical prescription data. Beyond that, the system has modules for appointment scheduling, prescription tracking, emergency-resource management, and centralised feedback handling. The system’s architecture is explained with the help of use-case activity sequence, data-flow, and deployment diagrams, and the system’s implementation is assessed via unit integration system, security, and user-acceptance testing, plus quantitative performance benchmarking and a comparative analysis of machine-learning models for the prescription recommendation task. The experimental evidence shows that the system developed is capable of matching the recommendation accuracy of baseline classifiers, while at the same time it greatly decreases administrative workload against traditional, isolated medical-record systems.

A Distributed Fog-Based Intrusion Detection Scheme for Industrial Internet of Things using Ensemble Approach

Anita Seth — Wed, 19 Aug 2026 00:00:00 +0000

The widespread growth of the Internet of Things (IoT) over the past few years has drastically transformed human life. Industrial IoT (IIoT) is a subset of the broader concept of IoT, which is predominantly deployed within the industrial environment. IIoT has revolutionized the industrial environment by automating business processes and enhancing production. However, this tremendous growth of IoT globally has greatly increased the threat of cyberattacks. Thus, there is a need to implement a lightweight, robust security system that can detect such attacks and take preventive measures. In this respect, the design of an intrusion detection scheme (IDS) gains importance in order to detect malicious activities and secure the data from misuse. In this work, a distributed intrusion detection scheme is proposed that involves the use of fog computing, dimensionality reduction, and an ensemble learning technique. The proposed model based on an ensemble learning algorithm is implemented on fog nodes, and the performance of the model has been analyzed using the Edge-IIoTset and Army Cyber Institute (ACI)-IoT2023 datasets. The results demonstrated the efficacy of the proposed model in terms of performance measures such as accuracy, detection rate, precision, and ROC (receiver operating characteristics).

Cloud Computing Adoption and Business Development in Small and Medium Enterprises in Bauchi State Nigeria

Olawale Surajudeen Adebayo, Suleiman Abba Bako — Mon, 24 Aug 2026 00:00:00 +0000

Cloud computing adoption among Nigerian SMEs remains low despite high awareness. This paper examines the adoption of cloud computing and business development among small and medium enterprises in Bauchi State, Nigeria. The study surveys 327 SMEs in Bauchi State. A structured questionnaire was used to collect primary data, comprising 6 sections covering demographics, awareness and perceptions, barriers, adoption status, operational efficiency, and competitiveness. Content validity was ensured through expert review and principal component analysis with varimax rotation confirmed the three-factor structure of the barrier items. Data were analyzed using R, while descriptive statistics were calculated for all variables. Hypotheses were tested using Chi-square, ordinal logistic regression, and Cohen’s d. Findings reveal 98.2% awareness but only 37.3% adoption. Security concerns (mean = 4.35), cost/infrastructure deficits (mean = 4.11), and lack of technical knowledge (mean = 4.02) were the main barriers. Awareness strongly correlated with adoption (χ² = 65.41, p < 0.001). Among adopters, higher barriers predicted greater efficiency (OR = 1.80 for cost/infrastructure, p = 0.012), but no competitiveness advantage was found between adopters and non-adopters (t = −0.16, p = 0.871). The study concludes that awareness alone does not translate to competitive advantage under infrastructural constraints.

Influence of Activation Functions in Deep Learning Models

B.J. Luckyn, Kizzy Elliot O, Idatonye Toby — Mon, 12 Oct 2026 00:00:00 +0000

Deep learning has become one of the most significant areas of artificial intelligence due to its ability to solve complex problems across various fields, including computer vision, natural language processing, healthcare, finance, and cybersecurity. A key component of deep learning models is the activation function, which introduces non-linearity into artificial neural networks and enables them to learn complex patterns from data. The choice of an activation function greatly influences a model’s learning capability, convergence speed, computational efficiency, and prediction accuracy. This study presents a comparative analysis of commonly used activation functions in deep learning models, including Sigmoid, Hyperbolic Tangent (Tanh), Rectified Linear Unit (ReLU), Leaky ReLU, Exponential Linear Unit (ELU), Swish, Gaussian Error Linear Unit (GELU), and Softmax. The study reviews their mathematical properties, advantages, limitations, and practical applications, as well as findings from previous research. In addition, the selected activation functions are compared based on performance metrics such as accuracy, convergence rate, training efficiency, and computational complexity. The findings indicate that while traditional activation functions such as Sigmoid and Tanh remain useful for specific applications, they are limited by issues such as the vanishing gradient problem. Modern activation functions, particularly ReLU and its variants, demonstrate improved performance in deep neural networks due to faster convergence and better computational efficiency. Furthermore, Swish and GELU have shown superior performance in several advanced deep learning applications. The study concludes that no single activation function is universally optimal, and the choice of an activation function should depend on the specific problem, model architecture, and application requirements.

Rule-Based Redaction and Verification for Structured Clinical Notes on IoMT Edge Gateways

Prashant D. Sawant — Thu, 15 Oct 2026 00:00:00 +0000

Small language models (SLMs) on hospital Internet of Medical Things (IoMT) edge gateways can receive identifiers, ignore SOAP or SBAR layouts, and add unsupported gendered titles or pronouns, yet no simple auditable control exists. The study proposes MedEdgePrompter, a four-stage, rule-based pipeline implemented using the Python standard library: redaction, template injection, model execution, and verification, followed by clinician review or flagging. Redaction replaces nine-digit Social Security Number-format numbers, medical record numbers, and dates of birth with placeholders; templates wrap text in SOAP or SBAR instructions. Verification checks header order and that gendered honorifics and pronouns in the output occur in the input. A scripted function replaced the model. Redaction was tested on 1,050 synthetic sentences and verification on 1,200 scripted outputs, with exact 95% Clopper-Pearson intervals. All 550 in-scope identifiers were redacted (CI 0.993 to 1.000). None of the 250 out-of-scope identifiers (names, telephone numbers, email addresses, street addresses, month-name dates) were redacted, whereas all 50 accession numbers were incorrectly redacted. The verifier flagged 500 of 500 injected faults versus 152 of 500 for baseline checks, but flagged helium “He” as a pronoun in all 100 cases. Guardrails took a median 0.077 ms per note (95th percentile 0.118 ms) on one vCPU, excluding model time. The rules are auditable but cover only defined patterns, do not satisfy HIPAA Safe Harbor, miss clinical errors, and need clinician review. Real SLM testing under clinical governance could support privacy-protective drafting, reduce documentation burden, and give auditable evidence for HIPAA, FDA, GDPR, EU MDR, and EU AI Act compliance.