Article Per Year (5 Year)
p-Index From 2021 - 2026
0.751
P-Index
This Author published in this journals
All Journal Journal of Information Systems and Informatics Emerging Information Science and Technology JEECS (Journal of Electrical Engineering and Computer Sciences) Journal of Technology Informatics and Engineering
Xin, Qi
Unknown Affiliation
Computer Science & IT Control & Systems Engineering Electrical & Electronics Engineering

Published : 4 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 4 Documents
Search

 1 
Hybrid Cloud Architecture for Efficient and Cost-Effective Large Language Model Deployment Xin, Qi
Journal of Information System and Informatics Vol 7 No 3 (2025): September
Publisher : Universitas Bina Darma

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.51519/journalisi.v7i3.1170

Abstract

Large Language Models (LLMs) have achieved remarkable success across natural language tasks, but their enormous computational requirements pose challenges for practical deployment. This paper proposes a hybrid cloud–edge architecture to deploy LLMs in a cost-effective and efficient manner. The proposed system employs a lightweight on-premise LLM to handle the bulk of user requests, and dynamically offloads complex queries to a powerful cloud-hosted LLM only when necessary. We implement a confidence-based routing mechanism to decide when to invoke the cloud model. Experiments on a question-answering use case demonstrate that our hybrid approach can match the accuracy of a state-of-the-art LLM while reducing cloud API usage by over 60%, resulting in significant cost savings and a ~40% reduction in average latency. We also discuss how the hybrid strategy enhances data privacy by keeping sensitive queries on-premise. These results highlight a promising direction for organizations to leverage advanced LLM capabilities without prohibitive expense or risk, by intelligently combining local and cloud resources.
Uncertainty-Aware Late Fusion for 3D Perception (Confidence Calibration + Fusion Rule Learning) Xin, Qi
Journal of Technology Informatics and Engineering Vol. 4 No. 1 (2025): APRIL | JTIE : Journal of Technology Informatics and Engineering
Publisher : University of Science and Computer Technology

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.51903/jtie.v4i1.485

Abstract

Late fusion remains attractive for multi-sensor 3D perception because it preserves independent sensor pipelines, enables modular upgrades, and supports rigorous ablation experiments. This paper presents an uncertainty-aware late-fusion framework that combines per-modality confidence calibration with learning a fusion rule. We conduct full experimental evaluations on a PandaSet-style LiDAR+camera subset comprising 10 multi-frame sequences and 2,200 synchronized frames, with 49,549 annotated 3D objects across the Car, Pedestrian, and Cyclist classes. The framework calibrates LiDAR and camera confidence using temperature scaling and isotonic regression, estimates uncertainty-conditioned localization variance, and fuses associated candidates using multiple rules (max, mean, product/odds, and Dempster-Shafer) as well as a learned fusion rule (logistic regression trained on association features). On the test split, isotonic calibration reduces LiDAR Expected Calibration Error from 0.260 to 0.006 and Negative Log-Likelihood from 0.410 to 0.110, and it similarly improves camera confidence quality. Although mean Average Precision (mAP) remains similar to a LiDAR-only baseline in this controlled setting, calibrated late fusion provides substantially better decision reliability at fixed confidence thresholds and maintains conservative high-precision behavior under camera dropout. These results support an engineering conclusion: confidence calibration is the highest-leverage upgrade for late fusion in safety-critical stacks, and fusion rule choice can be tuned to downstream risk preferences.
Self-Supervised Log Anomaly Detection with LogBERT-Style Transformers: Full Empirical Evaluation on a Reproducible SynHDFS Benchmark Xin, Qi
JEECS (Journal of Electrical Engineering and Computer Sciences) Vol. 11 No. 1 (2026): JEECS (Journal of Electrical Engineering and Computer Sciences) - In press
Publisher : Fakultas Teknik Universitas Bhayangkara

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.54732/jeecs.v11i1.3

Abstract

Log-based anomaly detection is a core problem in AIOps because system logs provide fine-grained evidence of failures, performance regressions, and security incidents. Recent work has shown that self-supervised sequence modeling substantially improves generalization compared with purely frequency-based detectors, especially when labeled anomalies are scarce. This paper presents a LogBERT-style transformer framework for session-level log anomaly detection and reports a complete, reproducible experimental evaluation. Due to download constraints of large archived log datasets in this environment, we construct a faithful fallback benchmark, SynHDFS-6k, which mimics HDFS-style block workflows by composing normal execution patterns and injecting five realistic anomaly types. SynHDFS-6k contains 6000 sessions with a fixed 5.0% anomaly rate and a vocabulary of 20 event templates. We train a two-layer transformer encoder with masked language modeling on normal sessions only and derive an anomaly score using pseudo log-likelihood (PLL) computed by masking each token position once. We compare against unigram and bigram probabilistic models, PCA reconstruction error, one-class SVM, isolation forest, a DeepLog-style GRU next-event predictor, and a supervised logistic regression upper bound. On the SynHDFS-6k test split, the proposed LogBERT-PLL achieves Precision=0.615, Recall=0.533, F1=0.571, ROC-AUC=0.898, and PR-AUC=0.594. We additionally analyze transformer scoring strategies (PLL mean, PLL top-k, PLL max, random masking, and CLS Mahalanobis), report runtime and model capacity trade-offs, and quantify per-anomaly-type detection behavior. The study provides an end-to-end blueprint for transformer-based self-supervised log anomaly detection under a fully specified protocol, and it highlights strengths and limitations that inform deployment on real-world HDFS logs.
Explaining OpenStack Failure-Injection Log Anomalies with Retrieved Normal Prototypes Xin, Qi
Emerging Information Science and Technology Vol. 6 No. 2 (2025)
Publisher : Universitas Muhammadiyah Yogyakarta

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.18196/eist.v6i2.31232

Abstract

OpenStack benchmark logs are widely used in log anomaly detection, but the official release also supports a more specific task: failure-injection-aware explanation. It provides normal logs and abnormal cases with injected failures, which makes it possible to evaluate anomaly ranking and case-based explanation on the same corpus. This paper reports a complete empirical evaluation of the official OpenStack files. We parsed 207,820 raw lines, constructed 2,070 file-scoped virtual-machine cases, and evaluated template, timing, and retrieval-based methods against the four officially labeled injected anomalies. The labeled failures are not characterized by anomaly-exclusive templates. Instead, they preserve the normal template vocabulary and are dominated by a large creating_image→vm_started delay, together with increased /servers/detail polling. A template-only Isolation Forest fails in this setting, achieving ROC-AUC 0.418 and F1@4 0.000, whereas timing-aware baselines reach ROC-AUC values between 0.978 and 0.997. We then introduce FIA-RAG, a deliberately simple and auditable explanation pipeline. FIA-RAG scores each case with three standardized benchmark-aligned signals case duration, detail-polling count, and create-to-start delay and retrieves nearby normal cases as prototypes for counterfactual repair. On the official abnormal file, it ranks all four labeled anomalies in the top four alerts and identifies the create-to-start delay as the dominant explanation signal. These results show that the official OpenStack failure cases are timing-and-context anomalies rather than lexical anomalies. More broadly, the study argues that failure-injection benchmarks should be characterized before model complexity claims are made. The three features used here are OpenStack-specific, but the protocol, case construction, prototype retrieval, and counterfactual validation, can be adapted to other cloud-log benchmarks after system-appropriate milestones and context counters are defined.
Co-Authors