ML_Proj

def predict(test_data):
    Test_df = pd.read_csv(test_data,header = 0, delimiter = ',', encoding="iso-8859-1")
    X_test = Test_df.text
    y_test = Test_df.IndustryCategory
    tokenizer = Tokenizer(num_words=5000)
    tokenizer.fit_on_texts(X_test)
    X_test = tokenizer.texts_to_sequences(X_test)
    maxlen = 100
    X_test = pad_sequences(X_test, padding='post', maxlen=maxlen)
    encoder = LabelEncoder()
    encoder.fit(y_test)
    y_test = encoder.transform(y_test)
    num_classes = np.max(y_test) + 1
    y_test = utils.to_categorical(y_test, num_classes)
    loaded_model=load_model('model_num.hdf5')
    loaded_model.compile(optimizer='adam',loss='binary_crossentropy',metrics=['accuracy'])
    loss, Test_accuracy = loaded_model.evaluate(X_test, y_test, verbose=False)
    prediction_new=loaded_model.predict_classes(X_test)
    prediction_class=encoder.inverse_transform(prediction_new)
    df_pre = pd.DataFrame({'prediction':prediction_class})
    Test_df=Test_df.reset_index(drop=True)
    df_out = pd.concat([Test_df,df_pre],axis=1)
    df_outt=df_out.drop(columns=['text'])
    pred=[]
    pred.append(loaded_model.predict_proba(X_test))
    precision=[]
    for i in pred:
        for x, y, z,n in i:
            precision.append(round(max(x, y,z,n),2))
    df_pre = pd.DataFrame({'precision':precision})
    df_final=pd.concat([df_outt,df_pre],axis=1)
    return (df_final.to_csv("category_prediction.csv", sep=',', encoding='iso-8859-1',index = False))