def get_data():
print(request)
area = request.args.get("area")
property_type = request.args.get("property")
rental_price = request.args.get("price")
surface_area = request.args.get("surface")
plot = request.args.get("plot")
query_input = []
for idx, var in enumerate([surface_area, rental_price, property_type]):
vars_query_input = [0] * len(data.all_var_types[idx])
idx_query_var = data.all_var_types[idx].index(var)
vars_query_input[idx_query_var] = 100
query_input.extend(vars_query_input)
#reshape query_input to correct format for input to our model
query_input = np.array(query_input).reshape(1, -1)
#retrain model based on new data
trained_model = models.train_model(data.model_data, data.area_names,
data.model_vars)
#have our trained model make a prediction based on our query input
_, probabilities = models.pred_proba(model=trained_model,
input_vars=query_input)
proba_idx = np.where(probabilities[0] == np.amax(probabilities[0]))
proba_idx = proba_idx[0][0]
#determine the index of the predicted area within the returned probabiblities array
pred_area = data.area_names[proba_idx]
proba = probabilities[0][proba_idx]
proba = '%.3f' % Decimal(proba)
#determine how the prediction probability of our previously predicted area has changed
#due to the change in data variables of this area
if area is not None:
proba_idx = data.area_names.index(area)
new_proba_prev_area = probabilities[0][proba_idx]
new_proba_prev_area = '%.3f' % Decimal(new_proba_prev_area)
plot_area = area
else:
new_proba_prev_area = None
plot_area = pred_area
del probabilities
if plot is not None:
plot_data = data.model_data.loc[data.model_data['area_name'] ==
plot_area]
plot_data = plot_data.loc[:, data.model_vars]
plot = plots.create_hbar(plot_area, plot_data)
return jsonify(prediction=pred_area,
prediction_proba=proba,
area_changed_proba=new_proba_prev_area,
plotData=plot)
else:
return jsonify(prediction=pred_area,
prediction_proba=proba,
area_changed_proba=new_proba_prev_area)
# Recommender demo by drinkredwine 2018
import app.models as m
import numpy as np
if __name__ == '__main__':
item_user_data = m.load_data('data/vi_dataset_events.csv')
model = m.train_model(item_user_data)
with open('data/vi_challenge_uID.csv', 'r') as f:
data = f.readlines()
for line in data:
userid = np.float(line)
recs = m.recommend(userid, d)
print(userid, recs)
def test_train(self):
data = load_data('data/user_items.csv')
train_model(data)