Business understanding - Kennu76/DataMiningProject GitHub Wiki

Background

Most modern-day delivery services rely on automated route optimization algorithms, that try to estimate the most optimal delivery routes, given the number of delivery vehicles, their payload weights and time constraints. While these algorithms can be mathematically very complex, they often discount road quality as one of the important conditional variables. Road quality is an important factor in route optimization, because it heavily affects the delivery time, fuel (or energy) consumption, and maintenance costs. In some cases, it may also affect the condition of the package, especially in case of fragile packages.

In real life, we know that road quality can vary greatly. We also know that road quality is not a static value, but it depends on weather conditions, traffic intensity, and seasonality. Additionally, these measurements have to be taken periodically to check whether the road quality has degraded or improved. So far, measuring the road quality in a dynamic manner has been too difficult to justify measurement costs. However, it has become more feasible with modern technology. Given gyroscope, accelerometer and GPS input from delivery vehicles, such as Starship delivery robots, we are now able to periodically measure road quality in an automated and relatively low-cost manner. This input can be used to enhance route optimization algorithms, save delivery time, lower maintenance costs, and improve client satisfaction.

Success criteria:

We consider our task successful, if we are able to measure road quality precisely and present our results in a format which allows it to be used for route optimization.

Data-mining goals:

  • Estimate a road-network based on Starship data
  • Measure road quality
  • Visualize road quality on a map

Business goals:

Integrate road quality with route optimization

Situation assessment

Resources: 3x analysts Starship data

Requirements and constraints:

Presentation must be ready by the 8th of January Must not share Starship data

Risks:

Initial 100GB dataset may be difficult to filter Since we do not have a guidebook, some features may be misunderstood Data may contain noise due to measurement errors which may influence the results

Terminology:

  • Accelometer
  • Gyroscope
  • Magnetometer
  • Delivery robot
  • Vehicle routing problem
  • Road (pavement) quality

Costs and benefits:

Estimated labor hours: 75 Benefit: Saved time, maintenance, and fuel costs, added customer satisfaction

Data mining goals

Goals: Road quality measurement Map Visualization of road qualities using google maps and leaflet

Data-mining success criteria

Since we have multiple measurements from a single geographical location, we can calculate statistically significant results concerning road quality. We will further validate our results using Google Streetview, by looking and the top 5 good road sections and 5 bad road sections visually (assuming that Streetview results are up to date and no road construction has been done in given locations since summer 2017). We will also try to contact Starship employees and ask if the results correspond with their experience.