duru juliet

Sentiment Analysis of E-Commerce Product-Based Reviews Using a Machine Learning Approach A Case Study of Amazon.com

With the rapid growth of e-commerce and online shopping platforms, product reviews have become a ... more With the rapid growth of e-commerce and online shopping platforms, product reviews have become a crucial resource for consumers making informed purchasing decisions. These reviews contain valuable insights into customer sentiments and opinions about products and services, significantly influencing other customers' choices. However, manually analysing large volumes of reviews is time-consuming and challenging. Machine learning offers an automated solution by extracting and classifying sentiments expressed in reviews. The primary objective of this research was to develop a sentiment analysis model using supervised machine learning to classify sentiments as positive or negative. The Amazon Fine Food dataset from Kaggle was utilised, comprising 568,454 reviews, from which a random sample of 5,685 reviews was drawn for analysis. Data preprocessing involved removing special characters and HTML tags, handling missing values, and eliminating stop words. Five supervised machine learning models were employed and compared: Logistic Regression, Random Forest, Decision Tree, Naive Bayes, and Support Vector Machine (SVM). The dataset was partitioned into training (75%), validation (25%), and test sets (20% of the total dataset). Models were evaluated using accuracy, precision, recall, F1-score, and classification reports. Logistic Regression achieved the highest accuracy (64.47%) and demonstrated the best overall performance. Aspect-based sentiment analysis was also explored to provide granular insights into specific product features. Results were visualised using word clouds, pie charts, bar charts, and heat maps. Logistic Regression was selected for deployment, with the model and TF-IDF vectorizer saved using the pickle library for efficient processing of new data. This research assists both consumers and businesses by facilitating informed decisions and enhancing product quality. Future work could explore deep learning architectures, advanced feature engineering, and resampling techniques such as SMOTE and ADASYN to address class imbalance.

Download

A wearable IoT-based health vital signs monitoring system to reduce mortality rate

Journal of Carcinogenesis, ISSN: 1477-3163 2025; Vol 24: Issue 3, https://carcinogenesis.com/ , 2025

Early detection and diagnosis of basic health condition will result in successful treatment of di... more Early detection and diagnosis of basic health condition will result in successful treatment of diseases and reduced fatality
rate in developing countries. It is possible to extract human bio-signals and use it to better understand the bodily health
status and reaction to external factors. This new Vital Smart System (VSS) design has in-built sensors which a user wears
as a wrist band without interfering with his/her daily life activities to monitor his/her major vital signs (BT, BP, HR), of
which current system has limitations to achieve. Using the Dynamic System Design Methodology (DSDM), a more
technically robust, cost-effective and portable real time device was designed. This device contains an IoT interface (BLE)
that connects to smart phones; a GSM interface (SIM900A) relying on AT command framework that accesses defined
medical personnel on the network in cases of emergencies; optical health LEDs that alerts the user; thus, enabling consistent
self-assessment of the individual’s health condition. In this research therefore, we provided a generalized smart three-in
one vital sign system that ensures effective monitoring and early detection of the user’s health status via an optimization
of offline and online cross-interfaces of standalone sensors.
Keywords: Real-time monitoring, Internet of Things (IoT), Temperature (BT), Vital signs, Wearable health device.

Download

A LIGHTWEIGHT CLIENT SIDE SENTIMENT ANALYSIS TOOL FOR REAL TIME PRODUCT REVIEW CLASSIFICATION

The rapid proliferation of e-commerce platforms and user-generated content has created an urgent ... more The rapid proliferation of e-commerce platforms and user-generated content has created an urgent demand for efficient, real-time sentiment analysis tools capable of processing large volumes of product reviews accurately and transparently. Existing solutions predominantly rely on server-side machine learning models or third-party cloud APIs, introducing concerns related to data privacy, response latency, infrastructure dependency, and limited interpretability. This paper presents a lightweight, browser-based sentiment analysis tool designed entirely for client-side execution, requiring no server infrastructure, API keys, or software installation. The proposed system employs a carefully curated lexicon-driven approach to classify product reviews as positive, negative, or neutral in real time, providing transparent explanations through the highlighting of influential sentiment-bearing words and the reporting of quantitative confidence scores. The system architecture comprises three integrated layers: a User Interface Layer for text input and result rendering, a Processing Layer implementing the JavaScript-based sentiment engine, and a Presentation Layer delivering animated, real-time visual feedback. The classification algorithm operates through a multi-stage text processing pipeline encompassing lowercasing, tokenisation, punctuation removal, and lexicon matching against two manually compiled word lists -one of 27 positive terms and one of 28 negative terms, derived from commonly occurring sentiment expressions in product review corpora. Sentiment polarity is determined by computing the difference between positive and negative word counts, while confidence is estimated using the ratio of sentiment-bearing words to total tokens, the magnitude of the polarity differential, and the presence of high-intensity sentiment indicators. Empirical evaluation using representative product reviews demonstrates that the tool achieves an end-to-end processing latency of under 100 milliseconds across all tested inputs, delivering instantaneous classification without perceptible delay. Three illustrative output scenariospositive (85% confidence), negative (75% confidence), and neutral (34% confidence) -are examined in detail, demonstrating the system's capacity to correctly identify and highlight sentiment-bearing tokens and to produce calibrated confidence distributions across polarity classes. The visual sentiment distribution interface further supports interpretability by providing users with an intuitive representation of the classification rationale. By operating exclusively within the client browser using HTML5, CSS3, and ES6 JavaScript, the proposed tool guarantees that no user data is transmitted to external servers, positioning it as a privacy-compliant solution aligned with GDPR and CCPA requirements. The system's dependency-free architecture makes it readily deployable in resource-constrained environments, including educational settings, small-scale ecommerce applications, and privacy-sensitive analytical workflows. Although lexicon-based approaches are inherently constrained in handling complex linguistic phenomena such as negation, sarcasm, and contextual polarity shifts -limitations acknowledged in the present work -the tool's design prioritises accessibility, transparency, and computational efficiency over raw classification accuracy. This research contributes a functional, open, privacy-preserving baseline for client-side sentiment analysis and identifies concrete directions for future enhancement, including negation and sarcasm detection, multilingual lexicon expansion, intensity scoring, and batch processing capabilities. The tool demonstrates that meaningful, interpretable sentiment assessment can be achieved without sacrificing user privacy, computational simplicity, or deployment accessibility.

Download

A Lightweight Client-Side Sentiment Analysis Tool for Real-Time Product Review Classification

Download

A Lightweight Client-Side Sentiment Analysis Tool for Real-Time Product Review Classification

The rapid proliferation of e-commerce platforms and user-generated content has created an urgent ... more The rapid proliferation of e-commerce platforms and user-generated content has created an urgent demand for efficient, real-time sentiment analysis tools capable of processing large volumes of product reviews accurately and transparently. Existing solutions predominantly rely on server-side machine learning models or third-party cloud APIs, introducing concerns related to data privacy, response latency, infrastructure dependency, and limited interpretability. This paper presents a lightweight, browser-based sentiment analysis tool designed entirely for client-side execution, requiring no server infrastructure, API keys, or software installation. The proposed system employs a carefully curated lexicon-driven approach to classify product reviews as positive, negative, or neutral in real time, providing transparent explanations through the highlighting of influential sentiment-bearing words and the reporting of quantitative confidence scores. The system architecture comprises three integrated layers: a User Interface Layer for text input and result rendering, a Processing Layer implementing the JavaScript-based sentiment engine, and a Presentation Layer delivering animated, real-time visual feedback. The classification algorithm operates through a multi-stage text processing pipeline encompassing lowercasing, tokenisation, punctuation removal, and lexicon matching against two manually compiled word lists -one of 27 positive terms and one of 28 negative terms, derived from commonly occurring sentiment expressions in product review corpora. Sentiment polarity is determined by computing the difference between positive and negative word counts, while confidence is estimated using the ratio of sentiment-bearing words to total tokens, the magnitude of the polarity differential, and the presence of high-intensity sentiment indicators. Empirical evaluation using representative product reviews demonstrates that the tool achieves an end-to-end processing latency of under 100 milliseconds across all tested inputs, delivering instantaneous classification without perceptible delay. Three illustrative output scenariospositive (85% confidence), negative (75% confidence), and neutral (34% confidence) -are examined in detail, demonstrating the system's capacity to correctly identify and highlight sentiment-bearing tokens and to produce calibrated confidence distributions across polarity classes. The visual sentiment distribution interface further supports interpretability by providing users with an intuitive representation of the classification rationale. By operating exclusively within the client browser using HTML5, CSS3, and ES6 JavaScript, the proposed tool guarantees that no user data is transmitted to external servers, positioning it as a privacy-compliant solution aligned with GDPR and CCPA requirements. The system's dependency-free architecture makes it readily deployable in resource-constrained environments, including educational settings, small-scale ecommerce applications, and privacy-sensitive analytical workflows. Although lexicon-based approaches are inherently constrained in handling complex linguistic phenomena such as negation, sarcasm, and contextual polarity shifts -limitations acknowledged in the present work -the tool's design prioritises accessibility, transparency, and computational efficiency over raw classification accuracy. This research contributes a functional, open, privacy-preserving baseline for client-side sentiment analysis and identifies concrete directions for future enhancement, including negation and sarcasm detection, multilingual lexicon expansion, intensity scoring, and batch processing capabilities. The tool demonstrates that meaningful, interpretable sentiment assessment can be achieved without sacrificing user privacy, computational simplicity, or deployment accessibility.

Download

Uploads

Papers by duru juliet

Log In