def test_home_rentals(self):
lightwood.config.config.CONFIG.USE_CUDA = USE_CUDA
lightwood.config.config.CONFIG.PLINEAR = PLINEAR
config = {
'input_features': [
{'name': 'sqft', 'type': 'numeric'},
{'name': 'days_on_market', 'type': 'numeric'},
{'name': 'neighborhood', 'type': 'categorical', 'dropout': 0.4}
],
'output_features': [
{'name': 'number_of_rooms', 'type': 'categorical', 'weights': {'0': 0.8, '1': 0.6, '2': 0.5, '3': 0.7, '4': 1}},
{'name': 'number_of_bathrooms', 'type': 'categorical', 'weights': {'0': 0.8, '1': 0.6, '2': 4}},
{'name': 'rental_price', 'type': 'numeric'},
{'name': 'location', 'type': 'categorical'}
],
'data_source': {'cache_transformed_data': CACHE_ENCODED_DATA},
'mixer': {
'class': NnMixer,
'kwargs': {
'selfaware': SELFAWARE,
'eval_every_x_epochs': 4,
'stop_training_after_seconds': 80
}
}
}
df = pd.read_csv('https://mindsdb-example-data.s3.eu-west-2.amazonaws.com/home_rentals.csv')
predictor = Predictor(config)
predictor.learn(from_data=df)
df = df.drop([x['name'] for x in config['output_features']], axis=1)
predictor.predict(when_data=df)
predictor.save('test.pkl')
predictor = Predictor(load_from_path='test.pkl')
for j in range(100):
pred = predictor.predict(when={'sqft': round(j * 10)})['number_of_rooms']['predictions'][0]
assert isinstance(pred, (str, int))
def run_full_test(USE_CUDA, CACHE_ENCODED_DATA, SELFAWARE, PLINEAR):
'''
Run full test example with home_rentals dataset
'''
lightwood.config.config.CONFIG.USE_CUDA = USE_CUDA
lightwood.config.config.CONFIG.PLINEAR = PLINEAR
config = {
'input_features': [{
'name': 'number_of_bathrooms',
'type': 'numeric'
}, {
'name': 'sqft',
'type': 'numeric'
}, {
'name': 'location',
'type': 'categorical'
}, {
'name': 'days_on_market',
'type': 'numeric'
}, {
'name': 'neighborhood',
'type': 'categorical',
'dropout': 0.4
}, {
'name': 'rental_price',
'type': 'numeric'
}],
'output_features': [{
'name': 'number_of_rooms',
'type': 'categorical',
# 'weights':{
# '0': 0.8,
# '1': 0.6,
# '2': 0.5,
# '3': 0.7,
# '4': 1,
# }
}],
'data_source': {
'cache_transformed_data': CACHE_ENCODED_DATA
},
'mixer': {
'class': lightwood.BUILTIN_MIXERS.NnMixer,
'selfaware': SELFAWARE
}
}
df = pd.read_csv(
"https://mindsdb-example-data.s3.eu-west-2.amazonaws.com/home_rentals.csv"
)
def iter_function(epoch, error, test_error, test_error_gradient,
test_accuracy):
print(
'epoch: {iter}, error: {error}, test_error: {test_error}, test_error_gradient: {test_error_gradient}, test_accuracy: {test_accuracy}'
.format(iter=epoch,
error=error,
test_error=test_error,
test_error_gradient=test_error_gradient,
accuracy=predictor.train_accuracy,
test_accuracy=test_accuracy))
predictor = Predictor(config)
# stop_training_after_seconds given in order to not get timeouts in travis
predictor.learn(from_data=df,
callback_on_iter=iter_function,
eval_every_x_epochs=4,
stop_training_after_seconds=40)
df = df.drop([x['name'] for x in config['output_features']], axis=1)
predictor.predict(when_data=df)
predictor.save('test.pkl')
predictor = Predictor(load_from_path='test.pkl')
preds = {}
for j in range(100):
pred = predictor.predict(
when={'sqft': round(j * 10)})['number_of_rooms']['predictions'][0]
if pred not in preds:
preds[pred] = 0
preds[pred] += 1
def iter_function(epoch, error, test_error, test_error_gradient, test_accuracy):
print(
'epoch: {iter}, error: {error}, test_error: {test_error}, test_error_gradient: {test_error_gradient}, test_accuracy: {test_accuracy}'.format(
iter=epoch, error=error, test_error=test_error, test_error_gradient=test_error_gradient,
accuracy=predictor.train_accuracy, test_accuracy=test_accuracy))
predictor.learn(from_data=data_frame, callback_on_iter=iter_function)
print('accuracy')
print(predictor.train_accuracy)
print('accuracy over all dataset')
print(predictor.calculate_accuracy(from_data=data_frame))
when = {'x': [1], 'y': [0]}
print('- multiply when. {when}'.format(when=when))
print(predictor.predict(when=when))
# saving the predictor
predictor.save('ok.pkl')
# loading the predictor
predictor2 = Predictor(load_from_path='ok.pkl')
when = {'x': [0, 0, 1, -1, 1], 'y': [0, 1, -1, -1, 1]}
print('- multiply when. {when}'.format(when=when))
print(predictor2.predict(when_data=pandas.DataFrame(when)))
when = {'x': [0, 3, 1, -5, 1], 'y': [0, 1, -5, -4, 7]}
print('- multiply when. {when}'.format(when=when))
print(predictor2.predict(when_data=pandas.DataFrame(when)))
def feedback(iter, error, test_error, test_error_gradient):
# predictor.stop_training()
print(
'iteration: {iter}, error: {error}, test_error: {test_error}, test_error_gradient: {test_error_gradient}, accuracy: {accuracy}'.format(
iter=iter, error=error, test_error=test_error, test_error_gradient=test_error_gradient,
accuracy=predictor.train_accuracy))
predictor.learn(from_data=data_frame, callback_on_iter=feedback)
print('accuracy')
print(predictor.train_accuracy)
print('accuracy over all dataset')
print(predictor.calculate_accuracy(from_data=data_frame))
when = {'x': [1], 'y': [0]}
print('- multiply when. {when}'.format(when=when))
print(predictor.predict(when=when))
# saving the predictor
predictor.save('/tmp/ok.pkl')
# loading the predictor
predictor2 = Predictor(load_from_path='/tmp/ok.pkl')
when = {'x': [0, 0, 1, -1, 1], 'y': [0, 1, -1, -1, 1]}
print('- multiply when. {when}'.format(when=when))
print(predictor2.predict(when_data=pandas.DataFrame(when)))
when = {'x': [0, 3, 1, -5, 1], 'y': [0, 1, -5, -4, 7]}
print('- multiply when. {when}'.format(when=when))
print(predictor2.predict(when_data=pandas.DataFrame(when)))
def run_test(USE_CUDA, CACHE_ENCODED_DATA, SELFAWARE, PLINEAR):
lightwood.config.config.CONFIG.USE_CUDA = USE_CUDA
lightwood.config.config.CONFIG.CACHE_ENCODED_DATA = CACHE_ENCODED_DATA
lightwood.config.config.CONFIG.SELFAWARE = SELFAWARE
lightwood.config.config.CONFIG.PLINEAR = PLINEAR
####################
config = {
'input_features': [{
'name': 'number_of_bathrooms',
'type': 'numeric'
}, {
'name': 'sqft',
'type': 'numeric'
}, {
'name': 'location',
'type': 'categorical'
}, {
'name': 'days_on_market',
'type': 'numeric'
}, {
'name': 'neighborhood',
'type': 'categorical',
'dropout': 0.4
}, {
'name': 'rental_price',
'type': 'numeric'
}],
'output_features': [{
'name': 'number_of_rooms',
'type': 'categorical',
# 'weights':{
# '0': 0.8,
# '1': 0.6,
# '2': 0.5,
# '3': 0.7,
# '4': 1,
# }
}],
'mixer': {
'class': lightwood.BUILTIN_MIXERS.NnMixer
}
}
# AX doesn't seem to work on the travis version of windows, so don't test it there as of now
if sys.platform not in ['win32', 'cygwin', 'windows']:
pass
#config['optimizer'] = lightwood.model_building.BasicAxOptimizer
df = pd.read_csv(
"https://mindsdb-example-data.s3.eu-west-2.amazonaws.com/home_rentals.csv"
)
def iter_function(epoch, error, test_error, test_error_gradient,
test_accuracy):
print(
'epoch: {iter}, error: {error}, test_error: {test_error}, test_error_gradient: {test_error_gradient}, test_accuracy: {test_accuracy}'
.format(iter=epoch,
error=error,
test_error=test_error,
test_error_gradient=test_error_gradient,
accuracy=predictor.train_accuracy,
test_accuracy=test_accuracy))
predictor = Predictor(config)
# stop_training_after_seconds given in order to not get timeouts in travis
predictor.learn(from_data=df,
callback_on_iter=iter_function,
eval_every_x_epochs=1,
stop_training_after_seconds=1)
predictor.save('test.pkl')
predictor = Predictor(load_from_path='test.pkl')
df = df.drop([x['name'] for x in config['output_features']], axis=1)
predictor.predict(when_data=df)
predictor.save('test.pkl')
predictor = Predictor(load_from_path='test.pkl')
preds = {}
for j in range(100):
pred = predictor.predict(
when={'sqft': round(j * 10)})['number_of_rooms']['predictions'][0]
if pred not in preds:
preds[pred] = 0
preds[pred] += 1