[Kaggle] London Bike, RNN and LSTM

https://www.kaggle.com/datasets/hmavrodiev/london-bike-sharing-dataset

London bike sharing dataset

 

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
%matplotlib inline

import os
for dirname, _, filenames in os.walk('/kaggle/input'):
    for filename in filenames:
        print(os.path.join(dirname, filename))

 

df = pd.read_csv('/kaggle/input/london-bike-sharing-dataset/london_merged.csv', parse_dates = ['timestamp'])
df.head()

 

df['timestamp']

 

df.shape

 

train = df.iloc[:17000, 1:2]
test = df.iloc[17000:17414, 1:2]

print(train.shape)
print(test.shape)

 

df['cnt'][:17000].plot(figsize = (15, 4), legend = True)
df['cnt'][17000:].plot(figsize = (15, 4), legend = True)
plt.legend(['train', 'test'])
plt.title('bike share demand')
plt.show()

 

from statsmodels.tsa.stattools import pacf

pacf = pacf(df['cnt'], nlags = 20, method = 'ols')
print(pacf)

from statsmodels.graphics.tsaplots import plot_pacf

plot_pacf(pacf, lags = 8, method = 'ols', title = 'pa').show

 

from sklearn.preprocessing import MinMaxScaler

sc = MinMaxScaler(feature_range = (0, 1))
train_scaled = sc.fit_transform(train)
train_scaled

 

X_train = []
y_train = []

for i in range(1, 17000):
    X_train.append(train_scaled[i-1:i, 0])
    y_train.append(train_scaled[i, 0])

X_train, y_train = np.array(X_train), np.array(y_train)
print(X_train.shape)

X_train = np.reshape(X_train, (X_train.shape[0], X_train.shape[1], 1))
print(X_train.shape)

 

from keras.models import Sequential
from keras.layers.core import Dense, Activation
from keras.layers.recurrent import SimpleRNN

 

rnn = Sequential()
rnn.add(SimpleRNN(activation = 'relu', units = 6, input_shape = (1, 1)))
rnn.add(Dense(activation = 'linear', units = 1))

print(rnn.summary())

 

rnn.compile(loss = 'mse', optimizer = 'adam', metrics = ['mse'])

 

rnn.fit(X_train, y_train, batch_size = 1, epochs = 2)

 

inputs = sc.transform(test)
inputs.shape

 

X_test = []

for i in range(1, 415):
    X_test.append(inputs[i-1:i,0])

X_test = np.array(X_test)
X_test = np.reshape(X_test, (X_test.shape[0], X_test.shape[1], 1))

 

rnn = rnn.predict(X_test)
rnn = sc.inverse_transform(rnn)

 

test1 = pd.DataFrame(test)
rnn1 = pd.DataFrame(rnn)

print(test1.shape)
print(rnn1.shape)

 

test1.plot(figsize = (15, 4), legend = True)
plt.legend(['cnt'])
plt.title('bike share demand')
plt.show()

 

rnn1.plot(figsize = (15, 4), legend = True)
plt.legend(['rnn'])
plt.title('bike share demand')
plt.show()

 

test = np.array(test)
test.shape

 

plt.figure(figsize = (15, 5))
plt.plot(test, marker='.', label='cnt', color='black')
plt.plot(rnn, marker=',', label='RNN', color='red')
plt.legend()

 

# LSTM

from keras.layers.recurrent import LSTM

lstm = Sequential()
lstm.add(LSTM(units = 6, activation = 'relu', input_shape = (1, 1)))
lstm.add(Dense(units = 1, activation = 'linear'))

 

print(lstm.summary())

 

lstm.compile(loss = 'mse', optimizer = 'adam', metrics = ['mse'])
lstm.fit(X_train, y_train, batch_size = 1, epochs = 2)

 

lstm = lstm.predict(X_test)
lstm = sc.inverse_transform(lstm)

 

plt.figure(figsize = (15, 5))
plt.plot(test, marker='.', label='cnt', color='black')
plt.plot(lstm, marker=',', label='LSTM', color='green')
plt.legend()