Hi$,$

I want to analyse ridership patterns using ML across various modes and learn how fare amount influences ridership.I want to know what feature engineering I should do for this and which techniques will help me in analysing. I have data on hourly and daily travel numbers across various modes. pls guide me on the implementation. Is the analysis better done over actual ridership numbers or percentages. why?

my data analysis till now based on ridership count is as follows:

$\%$change in ridership before and after fare changes:

Subway Bus

Bike Taxi

$8\mathrm{.}22536-5\mathrm{.}628627$

$16\mathrm{.}29588-8\mathrm{.}49553$

I have the following correlation matrix for before and after fare changes:

Subway Bus Bike Taxi

Subway $1\mathrm{.}0000000\mathrm{.}9743630\mathrm{.}2578130\mathrm{.}540510$

Bus $0\mathrm{.}9743631\mathrm{.}0000000\mathrm{.}3690970\mathrm{.}492550$

Bike $0\mathrm{.}2578130\mathrm{.}3690971\mathrm{.}0000000\mathrm{.}320243$

Taxi $0\mathrm{.}5405100\mathrm{.}4925500\mathrm{.}3202431\mathrm{.}000000$

Subway Bus Bike Taxi

Subway $1\mathrm{.}0000000\mathrm{.}9661800\mathrm{.}4480190\mathrm{.}452354$

Bus $0\mathrm{.}9661801\mathrm{.}0000000\mathrm{.}5559930\mathrm{.}374103$

Bike $0\mathrm{.}4480190\mathrm{.}5559931\mathrm{.}0000000\mathrm{.}286621$

Taxi $0\mathrm{.}4523540\mathrm{.}3741030\mathrm{.}2866211\mathrm{.}000000$