URGENT Traffic Prediction Model using AI/ML (LSTM)

Deliver by: July 27, 2025, 18:00 UTC (non-flexible)

I need Python developer to develop an AI/ML traffic prediction model using provided datasets. The project includes creating an artefact and a comprehensive 10,000-word report (cannot use generative AI). Please message me directly through chat.

The goal is to analyze temporal patterns and build an LSTM-based forecasting solution using the Department for Transport’s traffic count dataset.

• What will the future traffic volume be for a specific road segment in the UK? (Predict Annual Average Daily Flow (AADF) values for upcoming years using historical data.)

• What is the expected volume of different vehicle types (e.g., cars, HGVs, LGVs) for a given region or road segment? (regression models to estimate traffic distribution across vehicle types.)

• Which regions or roads are most at risk of high congestion due to heavy traffic? (Classify locations into risk categories (low/medium/high) using supervised learning.)

• Can we group/count points (roads) with similar traffic profiles to identify patterns across the UK? (clustering techniques to segment roads by usage, vehicle mix, and geography.)

• How do road features such as length, classification, and location influence traffic volume? (feature importance rankings to understand key drivers of traffic.)

• How does traffic differ between rural and urban roads in terms of vehicle mix and volume? (Compare traffic profiles using classification and statistical analysis.)

• Which road segments are likely to experience the highest growth in HGV traffic? (Forecast specific vehicle category trends to anticipate logistical load.)

• Can we estimate traffic for newly built or unmeasured road segments using nearby data? (spatial regression models or interpolation techniques for prediction.)

• Where should traffic sensors or enumeration points be placed next for optimal coverage? (AI recommendations for strategic sensor placement based on traffic patterns and gaps.)

• What will the total UK traffic volume by vehicle type look like in the next 5 years? (Aggregate regional forecasts to support national transport planning.)


Core Deliverables:

ML Artefact:
- Clean, modular, well-documented Python code implementing the LSTM traffic prediction model
- Use of Jupyter Notebook or Python scripts with comments and clear outputs
- Parameter tuning (e.g., lookback window, batch size, epochs, layers)

Comprehensive 10,000-Word Report:
- Problem statement and background
- Description of dataset and preprocessing methods
- Explanation of the LSTM model architecture and choices
- Model training and validation process
- Evaluation metrics (RMSE, MAE, R², etc.)
- Visualization of predictions vs actuals
- Discussion of training time and hardware used
- Limitations and future improvement suggestions


Ideal Skills & Experience:
- Expertise in Python and ML
- Experience with LSTM algorithms
- Strong documentation skills
- Ability to meet deadlines

Deliver by: July 27, 2025, 18:00 UTC (non-flexible)

Dataset: https://storage.googleapis.com/dft-statistics/road-traffic/downloads/data-gov-uk/dft_traffic_counts_raw_counts.zip
Dataset Definition: https://data.dft.gov.uk/gb-traffic-matrix/aadf-majorroads-metadata.pdf