def random_forest_cross_loc_time_step( timeStep, round1, location1, round2, location2, round3, location3, board ):
TIME_STEP = timeStep
rf = RandomForestRegressor()
data1 = loadMatrix( TIME_STEP, round1, location1, board)
data2 = loadMatrix( TIME_STEP, round2, location2, board)
cross = loadMatrix( TIME_STEP, round3, location3, board)
train = data1[0]
train = np.concatenate((train, data1[1]), axis=0 )
train = np.concatenate((train, data2[0]), axis=0 )
train = np.concatenate((train, data2[1]), axis=0 )
test = cross[1]
trainX = train[TIME_STEP::, NO2_INDEX]
trainX = np.expand_dims(trainX, axis=1)
trainX = np.concatenate( (trainX, np.expand_dims(train[TIME_STEP::, O3_INDEX], axis=1)), axis=1)
# trainX = np.concatenate( (trainX, np.expand_dims(train[TIME_STEP::, CO_INDEX], axis=1)), axis=1)
# trainX = np.concatenate( (trainX, np.expand_dims(train[TIME_STEP::, TEMP_INDEX], axis=1)), axis=1)
# trainX = np.concatenate( (trainX, np.expand_dims(train[TIME_STEP::, HUM_INDEX], axis=1)), axis=1)
for index in range(TIME_STEP):
trainX = np.concatenate( (trainX, np.expand_dims(train[TIME_STEP::, NO2_INDEX + (index + 1)*8], axis=1)), axis=1)
trainX = np.concatenate( (trainX, np.expand_dims(train[TIME_STEP::, O3_INDEX + (index + 1)*8], axis=1)), axis=1)
# trainX = np.concatenate( (trainX, np.expand_dims(train[TIME_STEP::, CO_INDEX + (index + 1)*8], axis=1)), axis=1)
# trainX = np.concatenate( (trainX, np.expand_dims(train[TIME_STEP::, TEMP_INDEX + (index + 1)*8], axis=1)), axis=1)
# trainX = np.concatenate( (trainX, np.expand_dims(train[TIME_STEP::, HUM_INDEX + (index + 1)*8], axis=1)), axis=1)
trainY = train[TIME_STEP::,EPA_NO2_INDEX]
trainY = np.expand_dims(trainY, axis=1)
trainY = np.concatenate( (trainY, np.expand_dims(train[TIME_STEP::, EPA_O3_INDEX], axis=1)), axis=1)
rf.fit(trainX, trainY)
testX = test[TIME_STEP::, NO2_INDEX]
testX = np.expand_dims(testX, axis=1)
testX = np.concatenate( (testX, np.expand_dims(test[TIME_STEP::, O3_INDEX], axis=1)), axis=1)
# testX = np.concatenate( (testX, np.expand_dims(test[TIME_STEP::, CO_INDEX], axis=1)), axis=1)
# testX = np.concatenate( (testX, np.expand_dims(test[TIME_STEP::, TEMP_INDEX], axis=1)), axis=1)
# testX = np.concatenate( (testX, np.expand_dims(test[TIME_STEP::, HUM_INDEX], axis=1)), axis=1)
for index in range(TIME_STEP):
testX = np.concatenate( (testX, np.expand_dims(test[TIME_STEP::, NO2_INDEX + (index + 1)*8], axis=1)), axis=1)
testX = np.concatenate( (testX, np.expand_dims(test[TIME_STEP::, O3_INDEX + (index + 1)*8], axis=1)), axis=1)
# testX = np.concatenate( (testX, np.expand_dims(test[TIME_STEP::, CO_INDEX + (index + 1)*8], axis=1)), axis=1)
# testX = np.concatenate( (testX, np.expand_dims(test[TIME_STEP::, TEMP_INDEX + (index + 1)*8], axis=1)), axis=1)
# testX = np.concatenate( (testX, np.expand_dims(test[TIME_STEP::, HUM_INDEX + (index + 1)*8], axis=1)), axis=1)
testY = test[TIME_STEP::, EPA_NO2_INDEX]
testY = np.expand_dims(testY, axis=1)
testY = np.concatenate( (testY, np.expand_dims(test[TIME_STEP::, EPA_O3_INDEX], axis=1)), axis=1)
y_predict = rf.predict(testX)
return y_predict, testY
def random_forest_time_step( timeStep, round, location, board ):
TIME_STEP = timeStep
rf = RandomForestRegressor()
data = loadMatrix( TIME_STEP, round, location, board)
train = data[0]
test = data[1]
trainX = train[TIME_STEP::, 1]
trainX = np.expand_dims(trainX, axis=1)
trainX = np.concatenate( (trainX, np.expand_dims(train[TIME_STEP::, 2], axis=1)), axis=1)
trainX = np.concatenate( (trainX, np.expand_dims(train[TIME_STEP::, 3], axis=1)), axis=1)
trainX = np.concatenate( (trainX, np.expand_dims(train[TIME_STEP::, 6], axis=1)), axis=1)
trainX = np.concatenate( (trainX, np.expand_dims(train[TIME_STEP::, 7], axis=1)), axis=1)
for index in range(TIME_STEP):
trainX = np.concatenate( (trainX, np.expand_dims(train[TIME_STEP::, 1 + (index + 1)*8], axis=1)), axis=1)
trainX = np.concatenate( (trainX, np.expand_dims(train[TIME_STEP::, 2 + (index + 1)*8], axis=1)), axis=1)
trainX = np.concatenate( (trainX, np.expand_dims(train[TIME_STEP::, 3 + (index + 1)*8], axis=1)), axis=1)
trainX = np.concatenate( (trainX, np.expand_dims(train[TIME_STEP::, 6 + (index + 1)*8], axis=1)), axis=1)
trainX = np.concatenate( (trainX, np.expand_dims(train[TIME_STEP::, 7 + (index + 1)*8], axis=1)), axis=1)
trainY = train[TIME_STEP::,4]
trainY = np.expand_dims(trainY, axis=1)
trainY = np.concatenate( (trainY, np.expand_dims(train[TIME_STEP::, 5], axis=1)), axis=1)
rf.fit(trainX, trainY)
testX = test[TIME_STEP::, 1]
testX = np.expand_dims(testX, axis=1)
testX = np.concatenate( (testX, np.expand_dims(test[TIME_STEP::, 2], axis=1)), axis=1)
testX = np.concatenate( (testX, np.expand_dims(test[TIME_STEP::, 3], axis=1)), axis=1)
testX = np.concatenate( (testX, np.expand_dims(test[TIME_STEP::, 6], axis=1)), axis=1)
testX = np.concatenate( (testX, np.expand_dims(test[TIME_STEP::, 7], axis=1)), axis=1)
for index in range(TIME_STEP):
testX = np.concatenate( (testX, np.expand_dims(test[TIME_STEP::, 1 + (index + 1)*8], axis=1)), axis=1)
testX = np.concatenate( (testX, np.expand_dims(test[TIME_STEP::, 2 + (index + 1)*8], axis=1)), axis=1)
testX = np.concatenate( (testX, np.expand_dims(test[TIME_STEP::, 3 + (index + 1)*8], axis=1)), axis=1)
testX = np.concatenate( (testX, np.expand_dims(test[TIME_STEP::, 6 + (index + 1)*8], axis=1)), axis=1)
testX = np.concatenate( (testX, np.expand_dims(test[TIME_STEP::, 7 + (index + 1)*8], axis=1)), axis=1)
testY = test[TIME_STEP::,4]
testY = np.expand_dims(testY, axis=1)
testY = np.concatenate( (testY, np.expand_dims(test[TIME_STEP::, 5], axis=1)), axis=1)
y_predict = rf.predict(testX)
#print(rf.get_params())
#print(rf.decision_path(testX))
return y_predict, testY
def random_forest_time_step(timeStep, round, location, board):
TIME_STEP = timeStep
nueral = NeuralNetwork(
['no2', 'o3', 'co', 'temperature', 'absolute-humidity', 'pressure'])
data = loadMatrix(TIME_STEP, round, location, board)
train = data[0]
test = data[1]
trainX = train[TIME_STEP::, 1]
trainX = np.expand_dims(trainX, axis=1)
trainX = np.concatenate(
(trainX, np.expand_dims(train[TIME_STEP::, 2], axis=1)), axis=1)
trainX = np.concatenate(
(trainX, np.expand_dims(train[TIME_STEP::, 3], axis=1)), axis=1)
trainX = np.concatenate(
(trainX, np.expand_dims(train[TIME_STEP::, 6], axis=1)), axis=1)
trainX = np.concatenate(
(trainX, np.expand_dims(train[TIME_STEP::, 7], axis=1)), axis=1)
for index in range(TIME_STEP):
trainX = np.concatenate(
(trainX,
np.expand_dims(train[TIME_STEP::, 1 + (index + 1) * 8], axis=1)),
axis=1)
trainX = np.concatenate(
(trainX,
np.expand_dims(train[TIME_STEP::, 2 + (index + 1) * 8], axis=1)),
axis=1)
trainX = np.concatenate(
(trainX,
np.expand_dims(train[TIME_STEP::, 3 + (index + 1) * 8], axis=1)),
axis=1)
trainX = np.concatenate(
(trainX,
np.expand_dims(train[TIME_STEP::, 6 + (index + 1) * 8], axis=1)),
axis=1)
trainX = np.concatenate(
(trainX,
np.expand_dims(train[TIME_STEP::, 7 + (index + 1) * 8], axis=1)),
axis=1)
trainY = train[TIME_STEP::, 4]
trainY = np.expand_dims(trainY, axis=1)
trainY = np.concatenate(
(trainY, np.expand_dims(train[TIME_STEP::, 5], axis=1)), axis=1)
nueral.fit(trainX, trainY)
testX = test[TIME_STEP::, 1]
testX = np.expand_dims(testX, axis=1)
testX = np.concatenate(
(testX, np.expand_dims(test[TIME_STEP::, 2], axis=1)), axis=1)
testX = np.concatenate(
(testX, np.expand_dims(test[TIME_STEP::, 3], axis=1)), axis=1)
testX = np.concatenate(
(testX, np.expand_dims(test[TIME_STEP::, 6], axis=1)), axis=1)
testX = np.concatenate(
(testX, np.expand_dims(test[TIME_STEP::, 7], axis=1)), axis=1)
for index in range(TIME_STEP):
testX = np.concatenate(
(testX,
np.expand_dims(test[TIME_STEP::, 1 + (index + 1) * 8], axis=1)),
axis=1)
testX = np.concatenate(
(testX,
np.expand_dims(test[TIME_STEP::, 2 + (index + 1) * 8], axis=1)),
axis=1)
testX = np.concatenate(
(testX,
np.expand_dims(test[TIME_STEP::, 3 + (index + 1) * 8], axis=1)),
axis=1)
testX = np.concatenate(
(testX,
np.expand_dims(test[TIME_STEP::, 6 + (index + 1) * 8], axis=1)),
axis=1)
testX = np.concatenate(
(testX,
np.expand_dims(test[TIME_STEP::, 7 + (index + 1) * 8], axis=1)),
axis=1)
testY = test[TIME_STEP::, 4]
testY = np.expand_dims(testY, axis=1)
testY = np.concatenate(
(testY, np.expand_dims(test[TIME_STEP::, 5], axis=1)), axis=1)
y_predict = nueral.predict(testX)
return y_predict, testY
import metasense
import data
from data import loadMatrix
import numpy as np
import matplotlib
matplotlib.use('Agg')
from matplotlib import pyplot as plt
data = loadMatrix(1, 'donovan', 19)
train = data[0]
train = train.sort_values(by=['datetime'])
mat = train['datetime'].values
mat = np.expand_dims(mat, axis=1)
mat = np.concatenate((mat, np.expand_dims(train['no2'].values, axis=1)),
axis=1)
mat = np.concatenate((mat, np.expand_dims(train['o3'].values, axis=1)), axis=1)
mat = np.concatenate((mat, np.expand_dims(train['co'].values, axis=1)), axis=1)
mat = np.concatenate((mat, np.expand_dims(train['epa-no2'].values, axis=1)),
axis=1)
mat = np.concatenate((mat, np.expand_dims(train['epa-o3'].values, axis=1)),
axis=1)
mat = np.concatenate(
(mat, np.expand_dims(train['temperature'].values, axis=1)), axis=1)
mat = np.concatenate(
(mat, np.expand_dims(train['absolute-humidity'].values, axis=1)), axis=1)
tempDates = np.chararray((len(train['datetime']), 1), itemsize=30)
for index in range(len(tempDates)):
if (index - 1) >= 0:
tempDates[index] = train['datetime'].values[index - 1]
else:
tempDates[index] = "no data"
def __init__(self, fileName):
self.data = data.loadMatrix(fileName)