identify and implement cluster approaches / variable designs

[GitTorres]

• function/purpose
• relationship
  • map out concurrent components
• simplicity
• consumer understanding

[GitTorres]

Proposed Rating Variable Structure

• (wip) - profile GM vehicles based on what components are installed

• (wip) function to map ADAS to category

  • category A - Autonomous Level
  • Autonomous level definitions: active, driving feedback, parking feedback, passive/convenience
  • category B - Intended Function (poc)

• (2nd) function to highlight co-installed components (threshold = 80%)

• (3rd) function to add vehicle counts

[GitTorres]

SAE reference: https://www.sae.org/news/2019/01/sae-updates-j3016-automated-driving-graphic

[GitTorres]

Deliverables:

• [x] FAQ

• [x] Table of install rates

  • [x] query grouped by model year, make, model, and sum(ADAS)
  • [x] in Python, group and compute install rates
  • [ ] in Python, create plots per group

• [x] Table of clusters (component, cluster)

  • [x] rows - ADAS, columns - cluster id, last column - recommended

• [ ] Initial discount recommendation

• [ ] Table of conditional install rates

  • [ ] For each my, get 19x19 table of conditional installation rates
  • [ ] in Python, create plot per row of table

[GitTorres]

No useful clustering ideas, because no 'features' of ADAS components from which we can group the components.

---

If the intent is to cluster a vehicle or a model:

1. Total # of ADAS components 1.1. For each vehicle V, let S be the total count of ADAS. 0 <= S <= 19. Input space is 1dim. Nice. 1.1.1. Cluster by model year and overall

2. ADAS component install rates 2.1. For each model M, let I_k be the install rate for ADAS component K. Input space is 19dim. 2.2. Cluster by model year and overall. Compare.