Building Recommender Systems with Machine Learning and AI

https://sundog-education.com/recsys/

Установили anaconda

https://www.anaconda.com/

$ conda install -c anaconda anaconda-navigator
$ conda install -c conda-forge scikit-surprise
$ anaconda-navigator

Запускаем Spyder

$ anaconda-navigator > Spyder

http://media.sundog-soft.com/RecSys/RecSys-Materials.zip

http://media.sundog-soft.com/RecSys/ml-latest-small.zip

Разархивизовать каталоги и ml-latest-small поместить в RecSys-Materials

GettingStarted/GettingStarted.py

03 Evaluating Recommender Systems

RUN --> Evaluating/TestMetrics.py

04 A Recommender Engine Framework

SVD - Singular Value Decomposition

RUN --> Framework/RecsBakeOff.py

Loading movie ratings...

Computing movie popularity ranks so we can measure novelty later...
Estimating biases using als...
Computing the cosine similarity matrix...
Done computing similarity matrix.
Evaluating  SVD ...
Evaluating accuracy...
Evaluating top-N with leave-one-out...
Computing hit-rate and rank metrics...
Computing recommendations with full data set...
Analyzing coverage, diversity, and novelty...
Computing the cosine similarity matrix...
Done computing similarity matrix.
Analysis complete.
Evaluating  Random ...
Evaluating accuracy...
Evaluating top-N with leave-one-out...
Computing hit-rate and rank metrics...
Computing recommendations with full data set...
Analyzing coverage, diversity, and novelty...
Computing the cosine similarity matrix...
Done computing similarity matrix.
Analysis complete.


Algorithm  RMSE       MAE        HR         cHR        ARHR       Coverage   Diversity  Novelty
SVD        0.9034     0.6978     0.0298     0.0298     0.0112     0.9553     0.0445     491.5768
Random     1.4385     1.1478     0.0089     0.0089     0.0015     1.0000     0.0719     557.8365

Legend:

RMSE:      Root Mean Squared Error. Lower values mean better accuracy.
MAE:       Mean Absolute Error. Lower values mean better accuracy.
HR:        Hit Rate; how often we are able to recommend a left-out rating. Higher is better.
cHR:       Cumulative Hit Rate; hit rate, confined to ratings above a certain threshold. Higher is better.
ARHR:      Average Reciprocal Hit Rank - Hit rate that takes the ranking into account. Higher is better.
Coverage:  Ratio of users for whom recommendations above a certain threshold exist. Higher is better.
Diversity: 1-S, where S is the average similarity score between every possible pair of recommendations
           for a given user. Higher means more diverse.
Novelty:   Average popularity rank of recommended items. Higher means more novel.

05 Content-Based Filtering

RUN --> ContentBased/ContentRecs.py

Loading movie ratings...

Computing movie popularity ranks so we can measure novelty later...
Estimating biases using als...
Computing the cosine similarity matrix...
Done computing similarity matrix.
Evaluating  ContentKNN ...
Evaluating accuracy...
Computing content-based similarity matrix...
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
of  8211
...done.
Analysis complete.
Evaluating  Random ...
Evaluating accuracy...
Analysis complete.


Algorithm  RMSE       MAE
ContentKNN 0.9375     0.7263
Random     1.4385     1.1478

Legend:

RMSE:      Root Mean Squared Error. Lower values mean better accuracy.
MAE:       Mean Absolute Error. Lower values mean better accuracy.

Using recommender  ContentKNN

Building recommendation model...
Computing content-based similarity matrix...
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
of  9066
...done.
Computing recommendations...

We recommend:
Presidio, The (1988) 3.841314676872932
Femme Nikita, La (Nikita) (1990) 3.839613347087336
Wyatt Earp (1994) 3.8125061475551796
Shooter, The (1997) 3.8125061475551796
Bad Girls (1994) 3.8125061475551796
The Hateful Eight (2015) 3.812506147555179
True Grit (2010) 3.812506147555179
Open Range (2003) 3.812506147555179
Big Easy, The (1987) 3.7835412549266985
Point Break (1991) 3.764158410102279

Using recommender  Random

Building recommendation model...
Computing recommendations...

We recommend:
Sleepers (1996) 5
Beavis and Butt-Head Do America (1996) 5
Fear and Loathing in Las Vegas (1998) 5
Happiness (1998) 5
Summer of Sam (1999) 5
Bowling for Columbine (2002) 5
Babe (1995) 5
Birdcage, The (1996) 5
Carlito's Way (1993) 5
Wizard of Oz, The (1939) 5