In [1]:
import splat
import wisps
import numpy as np
import matplotlib.pyplot as plt
from wisps import Annotator as an
from wisps import datasets
import pandas as pd
from tqdm import tqdm
%matplotlib inline
In [2]:
#scale the data
from sklearn.preprocessing import MinMaxScaler, StandardScaler
from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestClassifier
In [3]:
train_df=datasets['traing_set']
In [4]:
feats=['CH_4/H-Cont', 'CH_4/H_2O-1', 'CH_4/H_2O-2', 'CH_4/J-Cont',
'H-cont/H_2O-1', 'H-cont/H_2O-2', 'H-cont/J-Cont', 'H_2O-1/J-Cont',
'H_2O-2/H_2O-1', 'H_2O-2/J-Cont', 'spex_chi', 'snr2', 'snr1',
'line_chi', 'f_test', 'x', 'spt']
train_df['x']=train_df.spex_chi/train_df.line_chi
In [5]:
unc_columns=[x+'er' for x in wisps.INDEX_NAMES]
features=feats
In [6]:
prblm_feats=['line_chi', 'spex_chi', 'H_2O-2/J-Cont', 'H-cont/J-Cont', 'H_2O-1/J-Cont', 'H-cont/H_2O-1', 'snr2', 'x']
#pred_df[prblm_feats]=pred_df[prblm_feats].applymap(np.log10).replace(np.inf, np.nan).replace(-np.inf, np.nan).replace(np.nan, -999999.9)
train_df[prblm_feats]=train_df[prblm_feats].applymap(np.float).applymap(np.log10).replace(np.inf, np.nan).replace(-np.inf, np.nan).replace(np.nan, -999999.9)
In [7]:
def make_label_binary(labl):
##only two labels
if labl >0.:
return 1.
else:
return 0.
In [8]:
train_df['label']=train_df['label'].apply(make_label_binary)
In [9]:
from sklearn.metrics import confusion_matrix,accuracy_score
In [10]:
def compute_accuracy_score(features):
scaler = MinMaxScaler(feature_range=(0, 1))
#train_set=train_df[features]
X_train, X_test, y_train, y_test = train_test_split(train_df[features].values, train_df['label'].values, test_size=0.5, random_state=123456) ###grammar
scaler.fit(X_train)
X_train = scaler.transform(X_train)
X_test = scaler.transform(X_test)
rf = RandomForestClassifier(n_estimators=1000, oob_score=True, random_state=123456)
rf.fit(X_train, y_train)
pred_labels = rf.predict(X_test)
model_accuracy = accuracy_score(y_test, pred_labels)
return model_accuracy
In [11]:
scores=[]
fets=[]
for f in tqdm(features):
fets.append(f)
scores.append(compute_accuracy_score(fets))
In [12]:
from matplotlib.ticker import MultipleLocator
In [43]:
fealables=['idx'+ str(idx+1) for idx in range(0, len(wisps.INDEX_NAMES)) ]
fealbls=np.append(fealables, ['spexchi', 'snrj', 'snrh',
'linechi', 'ftest', 'x', 'spt'])
In [44]:
fealbls=[x.upper() for x in fealbls ]
In [46]:
fig, ax=plt.subplots(figsize=(6, 8))
plt.plot( scores, np.arange(len(features))+1)
ml = MultipleLocator(2.)
ml2=MultipleLocator(.05)
ax.xaxis.set_major_locator(ml2)
plt.grid(axis='y')
plt.yticks( np.arange(len(features))+1, fealbls, rotation='horizontal')
plt.ylabel('Features')
plt.xlabel('Accuracy')
plt.tight_layout()
plt.savefig(wisps.OUTPUT_FIGURES+'/cv_scores.pdf')
In [25]:
for x, y in zip(features, scores): print (x,y)
In [15]:
train_df.columns
Out[15]:
In [ ]: