TRISECT (Time-seRIeS forECasting Toolkit)

What is TRISECT

TRISECT is a simple and automated way to execute various deep learning models for time-series forecasting including:

• LSTM
• GRU
• BiDirectional LSTM
• CNN LSTM

---

Configuration File

In config.json file, several variables can be configured:

"epochs": ,
"batch_size": ,
"test_size": ,
"scaler": ,
"timestamp_column":,
"X":"",
"Y": "",
"num_past_samples": "",
"num_future_samples": ""

epochs: an arbitrary cutoff, generally defined as "one pass over the entire dataset", used to separate training into distinct phases, which is useful for logging and periodic evaluation

batch_size: the number of training examples in one forward/backward pass. The higher the batch size, the more memory space you'll need

train_size: this parameter decides the size of the data that has to be split as the train dataset

scaler: the scaler that will be used for the data. Either "MinMax" or "Standard" scaler can be used

timestamp_column: The csv files may sometimes contain columns related to the timestamp of each sample. Indicate the name of that column to remove it

X: Indicate the name of the columns that will be used as features. One or more columns can be used

Y: Indicate the name of the columns that the DL models need to forecast. One or more columns can be used

num_past_samples: Indicate how many time steps in the past the DL models need to take into account when forecasting

num_future_samples: Indicate how many time steps in the future the DL models need to forecast

---

An example cmd

python forecast.py -d daily-minimum-temperatures-in-me.csv -c config.json

After the execution of the desired model, the following plots are created and saved:

• Training/Test loss

Python version used

• Python 3.9.7 Interpreter