def get_notes(note_width=.25):
# if pickled result exists, return that
# exists = os.path.isfile('pickle/' + directory + '_notes')
# if exists:
# return pickle.load(open('pickle/' + directory + '_notes', 'rb'))
# if pickled result does not exist, create it and pickle it
piano = []
bass = []
sax = []
song_lens = []
songs = {}
num_songs = 0
for file in glob.glob(directory + "/*.mid"):
midi = None
try:
midi = converter.parse(file)
except Exception as e:
print('Could not parse file: {}'.format(file))
continue
print("Parsing %s" % file)
num_songs += 1
songs[file] = {}
instruments = instrument.partitionByInstrument(midi)
song_lens.append(midi.highestTime)
for i in instruments:
name = i.getInstrument().instrumentName
if name == 'Piano':
songs[file]['piano'] = i
piano.append(i)
elif name == 'Acoustic Bass':
songs[file]['bass'] = i
bass.append(i)
elif name == 'Saxophone':
songs[file]['sax'] = i
sax.append(i)
enumerated_notes, embedded = clean(songs, song_lens, note_width)
output = []
for s in embedded:
if len(s) == 3:
output.append(s)
pickle.dump(enumerated_notes, open('pickle/' + directory + '_encodings', 'wb'))
pickle.dump(output, open('pickle/' + directory + '_embedded', 'wb'))
return enumerated_notes, output
def load(grid_num):
df_train = pd.read_csv('train.csv',
delimiter=',',
parse_dates=True,
index_col="Unnamed: 0")
df_train.dataframeName = 'train.csv'
df_test = pd.read_csv("test.csv",
delimiter=",",
parse_dates=True,
index_col="Unnamed: 0")
df_test.dataframeName = 'test.csv'
x_train, y_train, x_test, y_test = clean_data.clean(
df_train, df_test, grid_num)
return x_train, y_train, x_test, y_test
def predict(test, xgb_model):
# with open('vec_model.pkl') as f:
# vec_model = pickle.load(f)
X_xgb = clean(test, isTrain=False, isjson=True)
# X_vec = clean_desc(test_file, isTrain=False)
xgb_preds = xgb_model.predict_proba(X_xgb)[:,1]
# vec_preds = vec_model.predict_proba(X_vec)[:,1]
# preds = (xgb_preds + vec_preds) / 2
# return X_xgb, X_vec, preds
return X_xgb, xgb_preds
# Load a potentially pretrained model from disk.
if datapath("model_small"):
lda_model = ldamodel.load(datapath("model_small"))
else:
# train model
lda_model = ldamodel(corpus, num_topics=10, id2word=dictionary)
pprint.pprint(lda_model.top_topics(corpus, topn=5))
# Save model to disk.
temp_file = datapath("model_small")
lda_model.save(temp_file)
return lda_model, corpus, dictionary
def visualize_pyldavis(lda_model, corpus, dictionary):
prepared = pyLDAvis.gensim.prepare(lda_model, corpus, dictionary)
pyLDAvis.save_html(prepared, 'vis_topic_model_02.html')
pyLDAvis.show(prepared)
if __name__ == '__main__':
#list of docs as lists of strings
texts = clean('voted-kaggle-dataset.csv')
lda_model, corpus, dictionary = train_model(texts)
# print(lda_model.show_topics())
visualize_pyldavis(lda_model, corpus, dictionary)
# CraigsRecommendation
# created by Mikaela Hoffman-Stapleton and Arda Aysu
from flask import Flask, request
import pandas as pd
import re
from datetime import datetime
from clean_data import clean
from filter_data import filter
from cluster_data import cluster
from gmaps_fns import *
from webpage import searchpage, recpage
# import json # for demo
pd.set_option('display.max_colwidth', -1)
listings = clean('craigslist.csv')
listings = listings.assign(index=range(len(listings)))
# for demo data:
# gmaps = pd.read_csv('gmaps.csv')
# listings = pd.merge(listings, gmaps, how='outer', on='index')
# with open('places_demo.json') as data_file:
# places = json.load(data_file)
# with open('distances_demo.json') as data_file:
# distances = json.load(data_file)
app = Flask(__name__)
@app.route('/')
def search_page():