route-quality-predictor

Machine Learning to predict the quality of a climbing route based on various attributes, images, and descriptors.

Binary classification

Note for any predictor, you can add -b to the end of the command line args to run in binary classification mode where a 3 or 4 star route is 1, and other routes are 0.

Create a lexicon stored as csv files for bag of word use.

python3 create_lexicon.py featuresInLexicon

Your lexicon will be saved in ./lexicons.

predict-svm.py

python3 predict-svm.py                # if you want all routes
python3 predict-svm.py Trad,Sport,TR  # if you want only these types of routes

predict-knn.py

python3 predict-knn.py 5      # 5 is k for knn

predict-bow-nn.py

python3 predict-bow-nn.py numberOfEpochs featuresInLexicon

Where featuresInLexicon in lexicon should correspond to a lexicon you created using the create_lexicon.py script. Your neural network model will be saved in ./models.

predict-bow-svm.py

python3 predict-bow-svm.py featuresInLexicon

predict-nn.py

python3 predict-nn.py numberOfEpochs

Your neural network model will be saved in ./models.

How to get route data.

# will create table and crawl, getting specified number of routes and parsing description out of html
python3 crawler.py max_num_routes
# will download the api json for each route that doesn't have it
python3 api.py
# will parse the api json for each route that has not been parsed
python3 api_parse.py

How to restore a neural net and run the "test" items in database on it.

Route description neural net. Make sure if the model was created in binary mode, you run this file also in binary mode.

python3 restore-bow-nn.py path_to_model
# ex:
python3 restore-bow-nn.py ./models/route_description_100_words_binary/model -b

Route features neural net.

python3 restore-nn.py path_to_model
# ex:
python3 restore-nn.py ./models/route_features_model_binary/model -b