notebook.community

Edit and run



In [1]:

    
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt



In [2]:

    
filenames = [
"7_Well239.xlsx","8_Well425.xlsx","9_Well445.xlsx","10_Well451.xlsx","11_Well496.xlsx",
"12_Well510.xlsx"
];
train_files = filenames[:5]



In [3]:

    
COLUMNS = ["DEPTH","RHOB","NPHI","GR","DT"]
FEATURES = ["RHOB","NPHI","GR"]



In [4]:

    
df = pd.DataFrame()



In [5]:

    
for item in train_files:
    cur_df = pd.read_excel('MTP/'+item, 'Sheet1', index_col=None)
    df = df.append(cur_df[COLUMNS],ignore_index=True)



In [6]:

    
print df.shape
df = df[(df.DT!=-999.25) & (df.RHOB!=-999.25) & (df.NPHI!=-999.25) & (df.GR!=-999.25)]
print df.shape









    



(14143, 5)
(8809, 5)



In [7]:

    
df.head(5)



In [8]:

    
df.head(5)



In [9]:

    
df = df.sort_values(by="DEPTH")



In [10]:

    
df.head(10)



In [11]:

    
base = df.iloc[0,0]
end = float(df.tail(1)["DEPTH"])



In [12]:

    
depth_coeff = []



In [13]:

    
from sklearn import linear_model

start = base
last_end = 0.0

while start < end:
    
    cur_df = df[(df.DEPTH<=(start+10)) & (df.DEPTH>(last_end))]
    
    x_train = cur_df[FEATURES]
    y_train = cur_df["DT"]
    
    clf = linear_model.LinearRegression(normalize=True)
    clf.fit(x_train,y_train)
    depth_coeff.append(clf)
    
    
    last_end = start
    start = start+10



In [14]:

    
def get_index(depth):
    index=-1
    start=base
    while depth>start:
        index = index+1
        start += 10
    
    return index

Testing



In [15]:

    
test_file = filenames[5]



In [16]:

    
test_file









    Out[16]:





'12_Well510.xlsx'



In [17]:

    
test_df = pd.read_excel('MTP/'+test_file, 'Sheet1', index_col=None)
test_df = test_df[COLUMNS]
test_df = test_df[(test_df.DT!=-999.25) & (test_df.RHOB!=-999.25) & (test_df.NPHI!=-999.25) & (test_df.GR!=-999.25)]



In [18]:

    
y_true = []
y_pred = []
for index,row in test_df.iterrows():
    
    cur_depth = row["DEPTH"]
    
    x_test = row[FEATURES]
    y_actual = row["DT"]
    
    y_true.append(y_actual)
    y_pred.append(clf.predict(x_test)[0])
    
    index = get_index(cur_depth)
    clf = depth_coeff[index]
    
#     print clf.predict(x_test),y_actual

R-squared metric



In [19]:

    
from sklearn.metrics import r2_score
r2_score(y_true, y_pred)









    Out[19]:





0.051253988808328654



In [20]:

    
rows = []
for clf in depth_coeff:
    row = []
    for item in clf.coef_:
        row.append(item)
    row.append(clf.intercept_)
    rows.append(row)



In [21]:

    
df = pd.DataFrame()
df = df.append(rows)
df.to_csv('coefficients.csv')



In [ ]:

	DEPTH	RHOB	NPHI	GR	DT
3	2775.5088	2.435110	0.346101	40.670404	98.957324
4	2775.6612	2.445212	0.347706	41.349963	99.668471
5	2775.8136	2.452603	0.352515	42.251214	100.067070
6	2775.9660	2.453797	0.361714	41.875382	100.234240
7	2776.1184	2.447196	0.371803	41.119991	100.398700

	DEPTH	RHOB	NPHI	GR	DT
3	2775.5088	2.435110	0.346101	40.670404	98.957324
4	2775.6612	2.445212	0.347706	41.349963	99.668471
5	2775.8136	2.452603	0.352515	42.251214	100.067070
6	2775.9660	2.453797	0.361714	41.875382	100.234240
7	2776.1184	2.447196	0.371803	41.119991	100.398700

	DEPTH	RHOB	NPHI	GR	DT
3535	2673.7056	2.360000	0.430734	64.326297	105.15000
3536	2673.8580	2.359511	0.437368	60.221711	105.15000
3537	2674.0104	2.323572	0.458727	62.227375	105.15000
3538	2674.1628	2.306771	0.489539	62.542708	105.15000
3539	2674.3152	2.304319	0.482252	61.706956	105.15000
3540	2674.4676	2.308664	0.459852	59.877785	105.84152
3541	2674.6200	2.264486	0.437299	58.674358	106.09804
3542	2674.7724	2.262016	0.423855	58.576253	103.32205
3543	2674.9248	2.273700	0.419237	55.232853	103.59746
3544	2675.0772	2.267772	0.424059	53.800148	103.55041