R functions (package TBD) to quicky find stories in data using data mining models.

These steps are working (ish):

clean: Clean the data. Right now this just means replacing Inf with NA. Will expand this over time.
dropnoisecols: Remove columns that cannot easily be analyzed or do not contribute information.
groupother: Group small groups.
dropoutliers: Remove outliers.
find_correlated_features: compare features to identify those that are correlated.
Method to easily plot correlated features.
Pick a target and identify which features drive it and the strength and direction of effects.

About Me

I'm an independent contractor helping companies build custom cloud apps and leverage data science, visual analytics, and AI. I offer low introductory rates, free consultation and estimates, and no minimums, so contact me today and let's chat about how I can help!

https://www.bryce-chamberlain.com/

Example

# packages required for storyteller.
require(glue)
require(magrittr)
require(progress)
require(reshape2)
require(ggplot2)

# easyr project setup.
require(easyr)
begin()

# read in some data. google or kaggle to find a dataset you are interested in. 
dt = read.any('myfile.ext')

# run the steps (functions in fun/ folder).
dt %<>% 
  clean(run_autotype = FALSE) %>% # read.any already runs autotype by default.
  dropnoisecols() %>%
  groupother() %>%
  dropoutliers() %>%
  correlatedfeatures_find()

# summarize patterns found in your data.
summary(dt)

# plot correlation between two variables.
plot_correlation(
  dt,  
  c('incident_date_year', 'age')
)

# fit a model against a target and identify key drivers.
dt %>%
  correlatedfeatures_address(
    target = 'total_claim_amount'
  ) %>%
  fitmodel(
    ignorecols = c('vehicle_claim', 'property_claim', 'injury_claim')
  ) %>%
  summary()

You can also visit https://www.kaggle.com/code/brycechamberlain/data-explore-automl/ for a notebook example.

superchordate/storyteller

About Me

Example