Skip to content
/ order-matters Public

An exploration of anchoring bias in machine learning labeling, particularly recommender systems

4 stars 1 fork Branches Tags Activity
Star
Notifications You must be signed in to change notification settings

atduskgreg/order-matters

BranchesTags

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

28 Commits
anchoring_experiment
anchoring_experiment
 
 
.gitignore
.gitignore
 
 
100k_multinomial.csv
100k_multinomial.csv
 
 
100k_observed_pair_distribution.csv
100k_observed_pair_distribution.csv
 
 
100k_pair_probabilities.csv
100k_pair_probabilities.csv
 
 
100k_user_sequences.csv
100k_user_sequences.csv
 
 
README.md
README.md
 
 
analysis.numbers
analysis.numbers
 
 
build_actual_pair_distribution.rb
build_actual_pair_distribution.rb
 
 
build_constant_predictor.rb
build_constant_predictor.rb
 
 
build_multinomial.rb
build_multinomial.rb
 
 
build_pair_probability.rb
build_pair_probability.rb
 
 
build_sequence_data.rb
build_sequence_data.rb
 
 
build_sequential_predictor.rb
build_sequential_predictor.rb
 
 
build_test_train.rb
build_test_train.rb
 
 
build_time_histogram.rb
build_time_histogram.rb
 
 
evaluate.rb
evaluate.rb
 
 
multinomial_pair_probability.rb
multinomial_pair_probability.rb
 
 
order_aware_mse.csv
order_aware_mse.csv
 
 
order_independent_mse.csv
order_independent_mse.csv
 
 
test.data
test.data
 
 
test.data.csv
test.data.csv
 
 
time_delta_histogram.csv
time_delta_histogram.csv
 
 
time_delta_histogram_zoom_zoomed.csv
time_delta_histogram_zoom_zoomed.csv
 
 
time_delta_histogram_zoomed.csv
time_delta_histogram_zoomed.csv
 
 
train.data
train.data
 
 
train.data.csv
train.data.csv
 
 
train_multinomial.csv
train_multinomial.csv
 
 
training_conditional_probabilities.csv
training_conditional_probabilities.csv
 
 
training_sequences.csv
training_sequences.csv
 
 

Repository files navigation

Order Matters

Code related to my current research work on accounting for anchoring bias on user labeling in machine learning systems

The code assumes you have the MovieLens 100k dataset. It loads the data in order to look for statistical evidence of influence of sequential recommendations.

TODO

  • Compare 100k_pair_probabilities.csv to 100k_observed_pair_distribution.csv to look for a statistically significant difference (pairwise t-test?)

  • Add some better explanation of how to reproduce the results. (Rake-based build script?)

  • Movie info about data file formats into a central place

About

An exploration of anchoring bias in machine learning labeling, particularly recommender systems

Resources

Readme
Activity

Stars

4 stars

Watchers

3 watching

Forks

1 fork
Report repository

Releases

No releases published

Packages

 
 
 

Languages