-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathmodelling_loop_v2.py
203 lines (147 loc) · 5.95 KB
/
modelling_loop_v2.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
import pandas as pd
import pylab as pl
import matplotlib.pyplot as plt
import numpy as np
from sklearn.ensemble import RandomForestRegressor
from sklearn.ensemble import ExtraTreesRegressor
from sklearn import cross_validation
from sklearn import datasets
from sklearn.metrics import r2_score
from sklearn.externals import joblib
from geotiffio import readtif
from geotiffio import createtif
from geotiffio import writetif
import os
import gc
import feature_extraction_tools_v2 as fe
# # import training data as data frame
# #data = pd.read_csv("D:/Julian/64_ie_maps/julian_tables_2/training_final_good.csv")
# data = pd.read_csv("D:/Julian/64_ie_maps/cleaning_training/train_ff3.csv")
# # replace -999 flags
# #data = data.replace("NA",np.nan)
# # 16 models must be made
# variables = [260,261,262,263,265,266,267,268,270,271,272,273,275,276,277,278]
# colnames = data.columns
# year_variable=284
# print("check year variable")
# print(colnames[year_variable])
# for i in range(len(variables)):
# target_variable=variables[i]
# print("check target variable")
# varname =colnames[target_variable]
# print(varname)
# # initialize model
# rf = RandomForestRegressor(n_estimators=1000,n_jobs=4,max_features=85,min_samples_split=5,oob_score=True)
# #rf = ExtraTreesRegressor(n_estimators=1000,n_jobs=4,max_features=30,min_samples_split=5,bootstrap=True,oob_score=True)
# # indices of variables of interest (target and covariates)
# selection = np.append([target_variable],range(4,258))
# selection = np.append(selection,[year_variable])
# # select data of interest
# data_selection = data.iloc[:,selection].as_matrix()
# # check type of array
# #print(np.dtype(data_selection))
# # force dtype = float32
# #data_selection = data_selection.astype(np.float32, copy=False)
# # complete cases
# data_selection = data_selection[~np.isnan(data_selection).any(axis=1)]
# data_selection = data_selection[np.isfinite(data_selection).any(axis=1)]
# #np.savetxt("foo.csv",data_selection, delimiter=",")
# # target variable / covariates
# y = data_selection[:,0].astype(np.float64)
# x = data_selection[:,1:]
# # split test-train
# #x_train, x_test, y_train, y_test = cross_validation.train_test_split(x,y, test_size=0.2, random_state=0)
# # fit model
# pmodel = rf.fit(x,y)
# path = "D:/Julian/64_ie_maps/models/"+varname+"/"
# if not os.path.exists(path):
# os.makedirs(path)
# #path="D:/Julian/64_ie_maps/models/average_tree_height/1000m/random_forest/ff3_rf_1000_85_5/"
# joblib.dump(pmodel,path+varname+'_ff3_rf_1000_85_5.pkl')
# # validation measures
# oobp = pmodel.oob_prediction_
# oobplog = np.exp(oobp)
# corrins = np.corrcoef(oobp,y)
# rmse = np.sqrt(np.mean((y - oobp)**2))
# mae = np.mean(np.absolute((y - oobp)))
# meanofresp = np.mean(y)
# # save model
# print(corrins)
# print(rmse)
# print(mae)
# print(meanofresp)
# statistics = np.array([corrins[0,1],rmse,mae,meanofresp])
# np.savetxt(path+"0"+varname+"statistics.csv",statistics, delimiter=",")
# # plot
# # Print the feature ranking
# # column names in searched_data_frame
# colnames = data.columns
# imp = pmodel.feature_importances_
# names = colnames[selection[1:]]
# imp,names = zip(*sorted(zip(imp,names)))
# index = range(len(names))
# columns = ['variable','importance']
# varimpdf = pd.DataFrame(index=index, columns=columns)
# varimpdf['variable']=names
# varimpdf['importance']=imp
# varimpdf = varimpdf.sort(columns="importance",ascending=False)
# varimpdf.to_csv(path+"0varimp_ff3_rf_1000_85_5.csv", sep=',', encoding='utf-8',index=False)
gc.collect()
# import training data as data frame
#data = pd.read_csv("D:/Julian/64_ie_maps/julian_tables_2/training_final_good.csv")
data = pd.read_csv("D:/Julian/64_ie_maps/modelling_20150702/training_tables_finales/final_train_20150716.csv")
colnames = data.columns
# 16 maps must be made
variables = [13]
for i in range(len(variables)):
target_variable=variables[i]
print("check target variable")
varname =colnames[target_variable]
print(varname)
# load model
#path = "D:/Julian/64_ie_maps/models/"+varname+"/"
path = "D:/Julian/64_ie_maps/modelling_20150702/models/"
pmodel = joblib.load(path+varname+'_c1c2_clean/treeheight20150716.pkl')
print(type(pmodel))
# load variable selection list:
imagesdf = pd.read_csv("D:/Julian/64_ie_maps/modelling_20150702/covariate_tables/covariates20150702.csv", header = 0)
# # filter raster
datasetf,rows,cols,bands = readtif("D:/Julian/64_ie_maps/rasters/filter/bov_cbz_km2.tif")
bandf = datasetf.GetRasterBand(1)
bandf = bandf.ReadAsArray(0, 0, cols, rows).astype(np.float32)
bandf = np.ravel(bandf)
# mexico body mask
baddatamask = bandf < 0
# testdata
nvar = int(len(imagesdf.index))
testdata = np.zeros(((cols*rows),nvar),dtype=np.float64)
for y in xrange(1):
year="2004_1"
print(year)
for i in xrange(len(imagesdf.index)):
# read images (variable of interest and associated quality product)
dataset,rows,cols,bands = readtif(imagesdf.iloc[i,2+y])
# make numpy array and flatten
band = dataset.GetRasterBand(1)
band = band.ReadAsArray(0, 0, cols, rows).astype(np.float32)
band = np.ravel(band)
maskmissings = (band <= -1.7e+308) | np.isnan(band) | np.isneginf(band) | np.isinf(band)
goodbandmean= np.mean(band[~maskmissings])
mask = maskmissings & (~baddatamask)
band[mask] = goodbandmean
band[baddatamask] = -1
testdata[:,i] = band
#testdata[:,nvar-1]=year
# remove empty cells
goodidx = testdata[:,0]!= -1
data = testdata[goodidx,:]
# prediction
print("predicting at last")
prediction = pmodel.predict(data)
predictionout = np.zeros((cols * rows),dtype=np.float64)
predictionout[predictionout==0]=-1
predictionout[goodidx]=prediction
outpath = "D:/Julian/64_ie_maps/modelling_20150702/products/"+varname+"_c1c2_clean/"
if not os.path.exists(outpath):
os.makedirs(outpath)
fe.save_file(dataset,predictionout, rows, cols, path=outpath, base_date=2004,month_base_date=1, varname=varname, sufix="please")