In [1]:

    

import pandas as pd
import numpy as np
import scipy
from sklearn import svm
import random


    

In [2]:

    

csvNames = ['Well1B3mths.csv','Well1C3mths.csv','Well1D3mths.csv','Well1E3mths.csv',
           'Well1F3mths.csv','Well1G3mths.csv','Well1H3mths.csv','Well1I3mths.csv',
           'Well1J3mths.csv','Well2A3mths.csv','Well2B3mths.csv','Well2C3mths.csv',
            'Well2D3mths.csv','Well2E3mths.csv','Well3A3mths.csv','Well3B3mths.csv',
            'Well3C3mths.csv','Well3D3mths.csv','Well3E3mths.csv','Well3F3mths.csv',
            'Well3G3mths.csv','Well3H3mths.csv','Well3I3mths.csv','Well4A3mths.csv',
            'Well4B3mths.csv']


    

In [3]:

    

featuresColNames = ['Casing Pressure',
                    'Gas Flow (Volume)',
                    'Motor Speed',
                    'Motor Torque',
                    'Pump Speed Actual', #this contains NULLs
                    'Tubing Flow Meter',
                    'Tubing Pressure',
                    'Water Flow Mag from Separator']
targetsName = ['Downhole Gauge Pressure']

allFeatures = []
allTargets = []

for well in csvNames:
    df = pd.DataFrame.from_csv(well)
    
    features = df[featuresColNames].dropna()
    target = df[targetsName].dropna()
    
    allFeatures.append(features)
    allTargets.append(target)

allFeatures = pd.concat(allFeatures).as_matrix()
allTargets = pd.concat(allTargets).as_matrix()


    

In [4]:

    

# Split into trainingX, trainingY, testX, testY

# shuffle these arrays so that training and test sets are randomized
print "shuffling the arrays"
allFeatures = random.sample(allFeatures, len(allFeatures)) 
allTargets = random.sample(allTargets, len(allTargets))

# percentage to train on
sampleSize = 0.1

trainSize = int(0.7*len(allFeatures)*sampleSize)
testSize = int(0.3*len(allFeatures)*sampleSize)
print "trainSize is: ", trainSize
print "testSize is: ", testSize

print "separating them into trainX, trainY, testX, testY"
trainX = allFeatures[:trainSize]
trainY = allTargets[:trainSize]
testX = allFeatures[trainSize:trainSize+testSize]
testY = allTargets[trainSize:trainSize+testSize]

print "done"


    

    




shuffling the arrays
trainSize is:  305640
testSize is:  130988
separating them into trainX, trainY, testX, testY
done

In [ ]:

    

clf = svm.SVR()
clf.fit(trainX, trainY) 
predictions = clf.predict(testX)


    

    




/usr/local/lib/python2.7/dist-packages/sklearn/utils/validation.py:526: DataConversionWarning: A column-vector y was passed when a 1d array was expected. Please change the shape of y to (n_samples, ), for example using ravel().
  y = column_or_1d(y, warn=True)

In [22]:

    

def calculateError(prediction, actual):
    return (predicted-actual)/actual

percentageErrors = []
for i, prediction in enumerate(predictions):
    actual = testY[i][0]
    percentageError = (prediction-actual)/actual
    percentageErrors.append(percentageError)

overallError = np.sum(percentageErrors) / len(percentageErrors)

print "Overall error was: " + str(overallError*100) + "%"


    

    




Overall error was: 51.8173081805%