Indoor localization using Wi-fi signal strength

• Pre-processing:

  • Normalization by row (not to overfit for phone models)
  • Excluding irregular values (outside of normal Wi-fi strength range)
  • From patterns in errors, exclude phone models that were trolling the system
  • Aggregate records of wi-fi strength in the same coordinate and average them -> Avoid volatily in Wi-fi strength

• Multi-level modelling -> Building (prediction) %>% Floor %>% Long & Lat (k-NN with 3 neighbours)

*Wi-fi pptx -> Presentation of the project to the management team