PREDICTIVE ANALYTICS AND DATA-DRIVEN ALGORITHMS FOR IMPROVING EFFICIENCY IN FULL-STACK WEB SYSTEMS
International Journal of Scientific Interdisciplinary Research · 01.06.2024 · 05 (02) · 226-260
DOI: 10.63125/q75tbj05 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Scientific Interdisciplinary Research, online erschienen am 01.06.2024. Verfasst von Md Muzahidul Islam und 1 weiteren Autor:innen. Die Arbeit wurde bisher 2-mal zitiert (Crossref).
Abstract (Originalsprache)
This study quantitatively examines how predictive analytics and data-driven algorithms improve the operational efficiency of full-stack web systems. Framed within the global expansion of cloud-native, containerized, and microservice architectures, we treat predictive components as embedded control mechanisms for load balancing, caching, and fault tolerance. Using a randomized post-test control design with longitudinal replication across a four-week period, concurrent user sessions were assigned to either a predictive optimization environment or a conventional reactive baseline. Telemetry captured request-level latency, throughput, CPU utilization, cache hit ratio, and model precision at one-minute intervals, under a standardized data quality protocol. Descriptively, the predictive system outperformed the baseline, yielding higher mean throughput (7,856 vs. 6,942 req/s, +13.1%), lower mean latency (182 ms vs. 247 ms, –26.3%), reduced CPU strain (68.3% vs. 72.8%), and improved cache efficiency (91.4% vs. 82.7%). Correlation analyses showed predictive precision was strongly associated with throughput (r = .81, p < .01) and inversely with latency (r = –.77, p < .01). A multiple regression model explained 74% of the variance in throughput (R² = .74; adjusted R² = .72), with predictive precision the dominant predictor (β = .62, p < .001) alongside cache hit ratio (β = .31, p < .01); workload intensity had a modest negative effect (β = –.18, p < .05). Logistic regression indicated systems with higher predictive precision were 2.8× more likely to sustain sub-200 ms latency. Model assumptions were satisfied (DW = 1.94), multicollinearity was acceptable (all VIFs < 3.5), and measurement reliability/validity were high (Cronbach’s α ≥ .89; CFA loadings ≥ .70). Findings establish that predictive precision is a statistically demonstrable driver of efficiency, stabilizing throughput and compressing tail latency via anticipatory resource control. We recommend operationalizing lightweight, hybrid predictive services within orchestration pipelines, instituting continuous feedback-driven recalibration, and adopting precision-linked SLOs to translate modeling gains into durable performance, scalability, and reliability improvements.
Zitieren
APA
Islam, M. M., & Dhanekula, A. (2024). PREDICTIVE ANALYTICS AND DATA-DRIVEN ALGORITHMS FOR IMPROVING EFFICIENCY IN FULL-STACK WEB SYSTEMS. International Journal of Scientific Interdisciplinary Research, 05(02), 226-260. https://doi.org/10.63125/q75tbj05
Vancouver
Islam MM, Dhanekula A. PREDICTIVE ANALYTICS AND DATA-DRIVEN ALGORITHMS FOR IMPROVING EFFICIENCY IN FULL-STACK WEB SYSTEMS. International Journal of Scientific Interdisciplinary Research. 2024;05(02):226-260. doi:10.63125/q75tbj05
BibTeX
@article{Islam2024,
author = {Islam, Md Muzahidul and Dhanekula, Aditya},
title = {PREDICTIVE ANALYTICS AND DATA-DRIVEN ALGORITHMS FOR IMPROVING EFFICIENCY IN FULL-STACK WEB SYSTEMS},
journal = {International Journal of Scientific Interdisciplinary Research},
year = {2024},
volume = {05},
number = {02},
pages = {226--260},
doi = {10.63125/q75tbj05}
}
EUVIMED