Logo image
Client-Side Predictive Component Pre-Rendering using Lightweight Machine Learning
Conference paper   Open access

Client-Side Predictive Component Pre-Rendering using Lightweight Machine Learning

Emmanuel Okang Edim, Salome Enoshi Uwah, Professor Celestine Iwendi, Vandana Sharma, Chukwuebuka Anthony Korie and Mubarak Olaniran Ibrahim
International Conference on Connected Intelligence for Industrial Applications (CI2A 2026) (Punjab, India, 03/04/2026–05/04/2026)
02/02/2026

Abstract

Index Terms—TensorFlowjs, Client-side Machine Learning, Predictive Rendering, Web Latency, Gated Recurrent Unit (GRU), Edge Computing
Contemporary web application architecture is under a severe bottleneck, the response time of reactive component rendering. With more dynamic user interfaces, user action to visual feedback latency is a major contributor to poor user experience (UX) and business conversion rates. In this study, the predictive model is proposed as a decentralised, client-side predictive framework, which uses Gated Recurrent Units (GRU) and is composed in TensorFlow.js, involving inference directly on the browser, the system is able to predict the navigation patterns of users and render the high probability items in advance. This approach leverages edge computing in order to maintain the privacy of data and decrease the load of the servers. The results of the experiment show that the perceived latency decreases by up to 87% on high-end devices without compromising on the prediction of accuracy of above 90%. So by deploying a GRU-based predictive framework in the browser, we found that an 83.2% intent prediction accuracy can be used to achieve a 79% perceived latency reduction on mobile devices. This research paper sets a scalable model of the anticipatory Web Design which predicts and responds to the needs of the user intuitively.
pdf
Client-Side Predictive Component Pre-Rendering using Lightweight Machine Learning1.62 MBDownloadView
AcceptedCC BY V4.0 Open Access
url
https://ci2a.info/View
Event Website

Metrics

9 File views/ downloads
21 Record Views

Details

Logo image

Usage Policy