TfL Northern Line Delay Analysis

This project explores London Underground performance using the TfL Lost Customer Hours dataset.
The aim is to analyse delay on the tube and compare it with other key lines (Central, Jubilee, Victoria).

---

🎯 Objectives

• Understand how Lost Customer Hours (LCH) evolve over time.
• Compare annual totals across multiple lines.
• Identify which periods and years drive the largest spikes in delays.
• Showcase data cleaning, analysis, and visualisation skills using Python (pandas, matplotlib, seaborn).

---

📊 Dataset

• Source: TfL London Underground Performance Data — London Datastore
• Measure: Lost Customer Hours (LCH) — the total time lost by passengers due to delays, weighted by passenger volumes.
• Granularity: 13 four-week Periods per Financial Year (FY).
• Coverage: Northern Line (focus), Central, Jubilee, and Victoria Lines (comparisons).
• Licence: Contains public sector information licensed under the Open Government Licence v3.0.

---

🛠️ Methods

1. Data Cleaning

   • Removed formatting artefacts (commas in numbers, blanks → NaN).
   • Converted all values to numeric.
   • Built a fy_period label (e.g., 2019-20 / P03) for readability.

2. Data Transformation

   • Period-level dataset cleaned and saved as /processed_data/period_clean.csv.
   • Annual totals aggregated and saved as /processed_data/annual_totals.csv.

3. Analysis & Visualisation

   • Multi-line trends across periods.
   • Rolling average trends (3-period smoothing).
   • Annual grouped bar charts.
   • Annual stacked bar charts (composition by line).
   • Heatmap experiment (evaluated but not selected for final portfolio).

---

📈 Key Visuals

1. Period Trends (Raw Data)

2. Rolling Average Trends

3. Annual Totals (Grouped Bars)

4. Annual Composition (Stacked Bars)

---

🔍 Insights

• The Central Line consistently shows the highest Lost Customer Hours across years.
• The Jubilee and Northern Lines show spikes in FY 2019–20, contributing a growing share to delays.
• Rolling averages reveal sustained upward trends prior to 2019–20, smoothing out period-to-period volatility.

---

🚀 Next Steps

• Interactive Dashboard: A Streamlit app allowing users to:
  • Select lines and year ranges.
  • Explore smoothed vs raw period trends.
  • Compare annual totals dynamically.
• Extend analysis: Add more lines and explore cross-modal comparisons (e.g., Overground, Buses).

---

⚖️ Licensing

• Data: © Transport for London, licensed under OGL v3.0.
• Code & Analysis: MIT Licence.

---

👤 Author

Built by Sarthak Nanda as part of a personal data analysis portfolio project.