def predict_and_submit(X_train, y_train, testFpath, clfKlass, **clfArgs):
X_test, testFnames = read_test_data_given_path(testFpath)
clf = clfKlass(**clfArgs)
clf.fit(X_train, y_train)
pred = np.zeros((X_test.shape[0], len(CLASS_NAMES)))
pred[:, np.sort(list(set(y_train)))] = clf.predict_proba(X_test)
make_submission_file(pred, testFnames, fNameSuffix='nonDN')
# X_train, y = read_train_data(width, height)
# x_fieldnames = np.array(['p_%i' % i for i in range(width*height)] + FEATURE_NAMES)
# #
# plot_feature_importances(fullX, ydata,
# np.array(['p'+str(i) for i in range(lastlayer.shape[1])] + FEATURE_NAMES),
# 0.7, numEstimators=100, min_samples_split=15)
# plot_pixel_importances(10, 20, lastlayer, ydata)
# for minSampleSplit in [17]:
# print minSampleSplit
# print evaluate(fullX, ydata, RandomForestClassifier,
# n_estimators=100, n_jobs=cpu_count()-1, min_samples_split=minSampleSplit)
featureVals_test = np.array(pandas.read_csv('/Users/jennyyuejin/K/NDSB/Data/X_test_48_48_featureVals.csv', header=None))
lastlayer_test = np.array(pandas.read_csv('/Users/jennyyuejin/K/NDSB/Data/lastlayerout_test.csv', header=None, sep=' '))
fullX_test = np.concatenate([lastlayer_test, featureVals_test], axis=1)
testFnames = list(np.array(pandas.read_table('/Users/jennyyuejin/K/NDSB/Data/testFnames.txt', header=None)).ravel())
clf = RandomForestClassifier(n_estimators=100, min_samples_split=20)
clf.fit(fullX, ydata)
pred = np.zeros((fullX_test.shape[0], len(CLASS_NAMES)))
pred[:, np.sort(list(set(ydata)))] = clf.predict_proba(fullX_test)
make_submission_file(pred, testFnames, fNameSuffix='withLastLayer')
