Telecom Network Operations & Fault Reduction Dashboard
End-to-end analytics pipeline analyzing fault patterns, MTTR, and infrastructure vulnerabilities with predictive risk scoring and an interactive Streamlit dashboard.
Simulation Note: This project uses a synthetic/sample dataset. Financial impact, savings, ROI, and SLA-cost estimates are modeled scenarios for analytical demonstration and do not represent actual NTC financial figures.
Modeled Estimate (synthetic data)
Interactive Dashboard
Fault Distribution by Type
MTTR by Fault Type (hours)
Faults by Technology
Cost Breakdown ($K)
Business Problem
High fault rates and long mean time to repair (MTTR) are costing the company in SLA penalties and customer churn. Which fault types, regions, and technologies (GPON vs Fiber) drive the most downtime and what actions will reduce MTTR the most?
Methodology
Data Generation
Generated synthetic telecom fault dataset with 2,000+ incidents across 4 connection types, 4 regions, and 12 months
Data Cleaning
Validated data types, handled missing values, standardized categorical fields, and created derived metrics for MTTR and fault rates
Analysis & Modeling
Performed fault frequency analysis, MTTR root-cause investigation, built Random Forest predictive model (82% accuracy), and ran statistical tests
Dashboard Deployment
Deployed interactive multi-page Streamlit dashboard with fault analysis, MTTR trends, cost simulation, and predictive risk scoring
Key Insights
Hardware/Power Fault Dominance
Hardware and Power faults account for 55% of all incidents but take 3x longer to resolve than other fault types --- primary targets for preventive maintenance.
GPON Infrastructure Vulnerability
GPON technology shows 35% of total faults, significantly higher than Fiber (20%), suggesting last-mile infrastructure aging that requires targeted upgrades.
MTTR and SLA Breach Risk
Average MTTR across all faults is 4.2 hours with an 18.5% SLA breach rate. Hardware faults average 6.8 hours and Power faults 8.2 hours.
Massive Cost Impact
Total annual downtime cost exceeds $2.8M with $420K+ in SLA penalties. Hardware/Power faults contribute 65% of total costs despite being only 55% of faults.
Predictive Risk Scoring
Random Forest model achieves 82% accuracy in predicting SLA breaches. Top 10% of sites are responsible for 40% of all faults.
Business Recommendations
Modeled Estimate (synthetic data)Prioritized Spare-Parts Stocking
Invest $125K for 50 high-fault sites to reduce Hardware/Power MTTR by 15%. Projected savings: $180K annually (144% ROI).
Preventive Maintenance Program
Invest $125K for 50 sites to reduce overall MTTR by 25%. Projected savings: $350K annually (280% ROI).
GPON Network Optimization
Targeted maintenance for GPON equipment to achieve 20% reduction in GPON-related faults.
Regional Focus
Increase support resources in North and East regions. Implement regional spare-parts depots to reduce response times.
Technical Highlights
Python & Pandas
- ▸Data cleaning and transformation
- ▸Time-series aggregation
- ▸Automated quality validation
Machine Learning
- ▸Random Forest classifier (82% accuracy)
- ▸SLA breach prediction
- ▸Feature importance ranking
Statistical Testing
- ▸Chi-Square for fault type vs SLA
- ▸ANOVA for MTTR across regions
- ▸Mann-Whitney U for technology comparison
Dataset
Synthetic telecom network fault dataset: 2,000+ incidents across 4 connection types, 4 regions, 12 months. Includes 400+ CIR/GPON connections.
Project Outcome
Modeled Estimate (synthetic data)This project demonstrates the ability to connect real telecom domain knowledge with advanced analytics techniques. The analysis identifies actionable infrastructure improvement opportunities with projected $530K+ annual savings through prioritized spare-parts stocking and preventive maintenance programs.
Limitations & Assumptions
Synthetic dataset - modeled business figures
All faults, incidents, MTTR values and costs were generated synthetically for this portfolio project. Downtime cost ($2.8M+), SLA penalties ($420K+) and projected savings ($530K+) are modeled outcomes from that synthetic data, not measured business results.
Accuracy reported as overall accuracy only
The 82% Random Forest score is overall accuracy on the synthetic dataset. No confusion-matrix breakdown, per-class precision/recall or F1 was computed; with class imbalance, overall accuracy can overstate real-world performance.
Cost simulation uses assumed parameters
Per-hour downtime cost and SLA penalty rates are assumed inputs to the simulation. Savings projections have not been sensitivity-tested across realistic cost ranges.
Single 12-month window
The data covers one synthetic 12-month period. Seasonality and year-over-year effects are illustrative and not validated against multiple periods.