EasyLSTM

EasyLSTM is a tool that eases the creation of LSTM's and shaping/slicing data to the correct shape.

With EasyLSTM you can go from import to prediction in less than 10 lines of python code!

Requirements

• TensorFlow >= 2.0
• Scikit-learn
• Keras
• Pandas
• Numpy

Usage

from easy_lstm import EasyLSTM

helper = EasyLSTM(data, timesteps)
model, X_train, y_train, X_test, y_test = helper.do_magic()

The do_magic() method returns both a compiled model and the train/test split and reshaped data.

The data goes through a series of processes that transform a pandas DataFrame into numpy arrays, with the X being an array of time windows of size n_steps from the dataframe feature columns, and the y being the corresponding labels in the dataset's ['y'] column.

Data format

The data provided to the EasyLSTM's __init__ method has to be a pandas dataframe in the following format:

The number of features/feature columns can be any number, as long as there is a at least one feature and a 'y' column

Sample DataFrame

Note: All feature columns can be named freely, just the label column that needs to be named 'y'

feature_0	feature_1	some_other_feature	y
0.1	2.12	1600	0.2
0.2	3.15	3000	0.3
0.3	3.98	400	0.4

7 Lines from import to prediction example

from easy_lstm import EasyLSTM
import pandas
dataset = pandas.read_csv('./path_to_dataset.csv')
dataset['y'] = dataset['feature'].shift(-1, axis=0)[:-1] #Turning a time series into a supervised learning problem
model, X_train, y_train, X_test, y_test = EasyLSTM(data=dataset, n_steps=4).do_magic()
model.fit(X_train, y_train, epochs=20)
predictions = model.predict(X_test)