import numpy as np
import pandas as pd
from matplotlib import pyplot as plt
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import LSTM
from tensorflow.keras.layers import Dense, Dropout
from sklearn.preprocessing import MinMaxScaler
from keras.wrappers.scikit_learn import KerasRegressor
from sklearn.model_selection import GridSearchCV

df=pd.read_csv("train.csv",parse_dates=["Date"],index_col=[0])
df.head()

df.tail()

df.shape
(5203,5)

test_split=round(len(df)*0.20)
df_for_training=df[:-1041]
df_for_testing=df[-1041:]
print(df_for_training.shape)
print(df_for_testing.shape)

(4162, 5)
(1041, 5)

scaler = MinMaxScaler(feature_range=(0,1))
df_for_training_scaled = scaler.fit_transform(df_for_training)
df_for_testing_scaled=scaler.transform(df_for_testing)
df_for_training_scaled

def createXY(dataset,n_past):
  dataX = []
  dataY = []
  for i in range(n_past, len(dataset)):
          dataX.append(dataset[i - n_past:i, 0:dataset.shape[1]])
          dataY.append(dataset[i,0])
  return np.array(dataX),np.array(dataY)

trainX,trainY=createXY(df_for_training_scaled,30)
testX,testY=createXY(df_for_testing_scaled,30)

data_X.addend (df_for_training_scaled[i - n_past:i, 0:df_for_training.shape[1]])

print("trainX Shape-- ",trainX.shape)
print("trainY Shape-- ",trainY.shape)

(4132, 30, 5)
(4132,)

print("testX Shape-- ",testX.shape)
print("testY Shape-- ",testY.shape)

(1011, 30, 5)
(1011,)

print("trainX[0]-- \n",trainX[0])
print("trainY[0]-- ",trainY[0])

trainX — — →trainY

[0 : 30,0:5] → [30,0]

[1:31, 0:5] → [31,0]

[2:32,0:5] →[32,0]

def build_model(optimizer):
  grid_model = Sequential()
  grid_model.add(LSTM(50,return_sequences=True,input_shape=(30,5)))
  grid_model.add(LSTM(50))
  grid_model.add(Dropout(0.2))
  grid_model.add(Dense(1))

grid_model.compile(loss = 'mse',optimizer = optimizer)
  return grid_modelgrid_model = KerasRegressor(build_fn=build_model,verbose=1,validation_data=(testX,testY))

parameters = {'batch_size' : [16,20],
            'epochs' : [8,10],
            'optimizer' : ['adam','Adadelta'] }

grid_search = GridSearchCV(estimator = grid_model,
                          param_grid = parameters,
                          cv = 2)

(trainX.shape[1],trainX.shape[2]) → (30,5)

grid_search = grid_search.fit(trainX,trainY)

grid_search.best_params_

{‘batch_size’: 20, ‘epochs’: 10, ‘optimizer’: ‘a(chǎn)dam’}

my_model=grid_search.best_estimator_.model

prediction=my_model.predict(testX)
print("prediction\n", prediction)
print("\nPrediction Shape-",prediction.shape)

scaler.inverse_transform(prediction)

prediction_copies_array = np.repeat(prediction,5, axis=-1)

prediction_copies_array.shape
(1011,5)

pred=scaler.inverse_transform(np.reshape(prediction_copies_array,(len(prediction),5)))[:,0]

original_copies_array = np.repeat(testY,5, axis=-1)
original=scaler.inverse_transform(np.reshape(original_copies_array,(len(testY),5)))[:,0]

print("Pred Values-- " ,pred)
print("\nOriginal Values-- " ,original)

plt.plot(original, color = 'red', label = 'Real Stock Price')
plt.plot(pred, color = 'blue', label = 'Predicted Stock Price')
plt.title('Stock Price Prediction')
plt.xlabel('Time')
plt.ylabel('Google Stock Price')
plt.legend()
plt.show()

df_30_days_past=df.iloc[-30:,:]
df_30_days_past.tail()

df_30_days_future=pd.read_csv("test.csv",parse_dates=["Date"],index_col=[0])
df_30_days_future

df_30_days_future["Open"]=0
df_30_days_future=df_30_days_future[["Open","High","Low","Close","Adj Close"]]
old_scaled_array=scaler.transform(df_30_days_past)
new_scaled_array=scaler.transform(df_30_days_future)
new_scaled_df=pd.DataFrame(new_scaled_array)
new_scaled_df.iloc[:,0]=np.nan
full_df=pd.concat([pd.DataFrame(old_scaled_array),new_scaled_df]).reset_index().drop(["index"],axis=1)

full_df_scaled_array=full_df.values
all_data=[]
time_step=30
for i in range(time_step,len(full_df_scaled_array)):
  data_x=[]
  data_x.append(
    full_df_scaled_array[i-time_step :i , 0:full_df_scaled_array.shape[1]])
  data_x=np.array(data_x)
  prediction=my_model.predict(data_x)
  all_data.append(prediction)
  full_df.iloc[i,0]=prediction

new_array=np.array(all_data)
new_array=new_array.reshape(-1,1)
prediction_copies_array = np.repeat(new_array,5, axis=-1)
y_pred_future_30_days = scaler.inverse_transform(np.reshape(prediction_copies_array,(len(new_array),5)))[:,0]
print(y_pred_future_30_days)