def pred(data, countries, start):
# ***** Get predictions for all country *****#
open('predictions.txt', 'w').close()
for country in countries:
record_predictions('\n' + country + ':')
print('\n', country, ':')
df = data[['Week', country]]
df = df[np.isfinite(df[country])] # remove empty rows
df = df.rename(index=str, columns={country: "Now"})
for i in range(3):
df_ = df.copy()
df_ = prepare(df_, pred=True)
# train the model using all data except for the latest record
X = df_.iloc[:,:-1]
y = df_.iloc[:,-1]
X,y = model.preprocess_data(X,y)
X_train, y_train = X[:-1], y[:-1]
# printX_train, '\n', y_train
lr, label = model.lr(X_train,y_train)
model.evaluate(lr,X,y,X_train,X_train,y_train,y_train,label,graph=False)
fea = X[-1]
y_pred = lr.predict([fea])
print('features:', fea, '\npredictions:', y_pred[0])
record_predictions(str(y_pred[0]))
rec = pd.DataFrame([[start+i, y_pred[0]]], columns=['Week','Now'])
# print'\nbefore appending\n', df
df = df.append(rec)
def index():
"""
POSTでリクエストを受け付けた場合、画像をtensorflowで分析して予測結果を返す
"""
if request.method == 'GET':
return render_template('index.html')
# NOTE: POST request
file = request.files['file']
file_body = file.read()
# tensorflowに読み込ませるように画像を変換
image_data = preprocess_data(
255 -
np.array(Image.open(BytesIO(file_body)).resize((28, 28)).convert('L')))
# TODO: 毎回modelを訓練するのではなく、保存したものを使い回すようにする
# NOTE: なぜかトップレベルでmodelを定義すると"Tensor(...) is not an element of this graph."というエラーが出る
model = build_model()
# 判定
prediction = model.predict(np.array([image_data]))[0]
# 小数点2桁で四捨五入
prediction = [round(float(n), 4) for n in prediction]
print(prediction)
return render_template(
'index.html',
image_source=base64.b64encode(file_body).decode(
"utf-8"), # 画像の保存があるとなにかと面倒くさいのでbase64で
prediction=sorted(dict(zip(class_names_ja, prediction)).items(),
key=lambda x: x[1],
reverse=True),
)
def runTests():
model.load_data('data/train.csv', 'data/test.csv')
methodPreprocess_list = [1, 2]
methodFeatureExtraction = [1, 2]
maxFeatures_list = [16000, 20000, 25000]
ngrams_list = [(1, 3), (1, 4), (1, 5), (1, 2)]
maxdf_list = [1.0, 0.95, 0.9, 0.85]
mindf_list = [0.0001]
binary_list = [True]
for methodPreprocess in methodPreprocess_list:
train_tweets, test_tweets = model.preprocess_data(
method=methodPreprocess)
for maxfeat in maxFeatures_list:
for ng in ngrams_list:
for maxd in maxdf_list:
for mind in mindf_list:
for bin in binary_list:
for featureExt in methodFeatureExtraction:
X, y, test, feature_names = model.feature_extraction(
train_tweets,
test_tweets,
maxFeatures=maxfeat,
ngrams=ng,
maxdf=maxd,
mindf=mind,
isbinary=bin,
method=featureExt)
print('\n\n\n')
print(
"#########################################################################"
)
print(
"##############################################"
)
print(
'Params preprocessing and features extraction:'
)
print(
'{0}, {1}, {2}, {3}, {4}, {5}, {6}'.format(
methodPreprocess, maxfeat, ng, maxd,
mind, bin, featureExt))
print(
"##############################################"
)
gsearch(X, y)
print(
"#########################################################################"
)
print('\n\n\n')
def second():
text = [x for x in request.form.values()]
processed_data = preprocess_data(text, tokenizer)
results = new_model.predict(processed_data)
print(results)
if results > .75:
positive = results
negative = 1 - positive
else:
positive = results
negative = 1 - positive
return render_template('index.html', positive=positive, negative=negative)
@author: Ashima
"""
import config
#import data
import model
import pandas as pd
import numpy as np
import seaborn as sns
from sklearn.datasets import load_wine
if __name__ == "__main__":
#Load Data
data = load_wine(return_X_y = False)
#build model
model = model.Model(data)
model.preprocess_data()
#model.visualize_data()
# #Train using SVM CLassifier_One Vs One
#model.linearsvc_ovo()
#print("SVM trained model-One Vs One")
## #Train using SVM CLassifier_One Vs One
#model.linearsvc_ovr()
#print("SVM trained model-One Vs Rest")
# Train using Gaussian Naive Bayes Classifier
#model.naive_bayes()
#print("Naive Bayes trained model")
#Train using Decision trees Classifier
model.decision_trees()
#print("Decision Trees trained model")
from scipy.sparse import coo_matrix, hstack
#
#model.load_data('data/train.csv', 'data/test.csv')
#train_tweets, test_tweets = model.preprocess_data(method=1)
#dictFeatures_train, dictFeatures_test, y = model.new_extractFeatures(train_tweets, test_tweets)
#results, s = model.new_gsearchRidge(dictFeatures_train, dictFeatures_test, y, nSplits = 3, testSize=0.4)
#
#
#
#clfs = model.new_train3(dictFeatures_train, y)
#train_prediction, l = model.new_predict3(clfs, dictFeatures_train, y=y)
#test_prediction, l = model.new_predict3(clfs, dictFeatures_test, y=None, clfsLabels=l)
#model.saveResults('output/xxx1.csv', test_prediction)
model.load_data('data/train.csv', 'data/test.csv')
train_tweets, test_tweets = model.preprocess_data(method=1)
emo_train, emo_test, sent_train, sent_test, k_train, k_test, w_train, w_test = model.new_extract_addicional_features(
)
l_tfidf = []
l_cv = []
print()
print()
for mF in [
12500, 15000, 20000, 25000, 30000, 35000, 40000, 45000, 50000, 55000,
60000
]:
print('Max Features = {0}'.format(mF))
for maxd in [0.55, 0.5, 0.45, 0.4, 0.35, 0.3, 0.25, 0.2]:
print(
for i in range(min([len(images), 25])):
plt.subplot(5, 5, i + 1)
plt.xticks([])
plt.yticks([])
plt.grid('off')
plt.imshow(images[i], cmap=plt.cm.binary)
print(predictions[i])
predicted_label = np.argmax(predictions[i])
true_label = labels[i]
if predicted_label == true_label:
color = 'green'
else:
color = 'red'
accuracy = round(float(predictions[i][predicted_label]), 4)
label = "{} ({}) ({})".format(class_names[predicted_label],
class_names[true_label], accuracy)
plt.xlabel(label, color=color)
plt.savefig('test_results/{}.png'.format(datetime.now().strftime('%s')))
plt.show()
_, (test_images, test_labels) = keras.datasets.fashion_mnist.load_data()
test_images = preprocess_data(test_images)
model = build_model()
run(model, test_images, test_labels)
from model import preprocess_data
from model import keras_model
from model import simple_model
from keras.utils.np_utils import to_categorical
from keras import optimizers
size = 64
class_size = 3500
batch_size = 32
X, y = load_data(size, class_size)
n_classes = len(np.unique(y))
print "Number of classes =", n_classes
print "Img set shape", X.shape
X = preprocess_data(X)
y = to_categorical(y)
print "y shape", y.shape
rand_state = np.random.randint(0, 100)
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=0.2,
random_state=rand_state)
print "train shape X", X_train.shape
print "train shape y", y_train.shape
# np.save("X_train.data", X_train)
# np.save("y_train.data", y_train)
# np.save("X_test.data", X_test)
# np.save("y_test.data", y_test)
