trans
Perform Transformer Forecast: https://unit8co.github.io/darts/generated_api/darts.models.forecasting.transformer_model.html
Usageโ
trans [--d-model D_MODEL] [--nhead NHEAD] [--num_encoder_layers NUM_ENCODER_LAYERS] [--num_decoder_layers NUM_DECODER_LAYERS] [--dim_feedforward DIM_FEEDFORWARD] [--activation {relu,gelu}] [--past-covariates PAST_COVARIATES] [--all-past-covariates] [--naive] [-d {AAPL}] [-c TARGET_COLUMN] [-n N_DAYS] [-t TRAIN_SPLIT] [-i INPUT_CHUNK_LENGTH] [-o OUTPUT_CHUNK_LENGTH] [--force-reset FORCE_RESET] [--save-checkpoints SAVE_CHECKPOINTS] [--model-save-name MODEL_SAVE_NAME] [--n-epochs N_EPOCHS] [--dropout DROPOUT] [--batch-size BATCH_SIZE] [--end S_END_DATE] [--start S_START_DATE] [--residuals] [--forecast-only] [--export-pred-raw] [--metric {rmse,mse,mape,smape}]
Parametersโ
Name | Parameter | Description | Default | Optional | Choices |
---|---|---|---|---|---|
d_model | --d-model | Number of expected features in inputs. | 64 | True | None |
nhead | --nhead | Number of head in the attention mechanism. | 4 | True | None |
num_encoder_layers | --num_encoder_layers | The number of encoder layers in the encoder. | 3 | True | None |
num_decoder_layers | --num_decoder_layers | The number of decoder layers in the encoder. | 3 | True | None |
dim_feedforward | --dim_feedforward | The dimension of the feedforward model. | 512 | True | None |
activation | --activation | Number of LSTM layers. | relu | True | relu, gelu |
past_covariates | --past-covariates | Past covariates(columns/features) in same dataset. Comma separated. | None | True | None |
all_past_covariates | --all-past-covariates | Adds all rows as past covariates except for date and the target column. | False | True | None |
naive | --naive | Show the naive baseline for a model. | False | True | None |
target_dataset | -d --dataset | The name of the dataset you want to select | None | True | AAPL |
target_column | -c --target-column | The name of the specific column you want to use | close | True | None |
n_days | -n --n-days | prediction days. | 5 | True | None |
train_split | -t --train-split | Start point for rolling training and forecast window. 0.0-1.0 | 0.85 | True | None |
input_chunk_length | -i --input-chunk-length | Number of past time steps for forecasting module at prediction time. | 14 | True | None |
output_chunk_length | -o --output-chunk-length | The length of the forecast of the model. | 5 | True | None |
force_reset | --force-reset | If set to True, any previously-existing model with the same name will be reset (all checkpoints will be discarded). | True | True | None |
save_checkpoints | --save-checkpoints | Whether to automatically save the untrained model and checkpoints. | True | True | None |
model_save_name | --model-save-name | Name of the model to save. | trans_model | True | None |
n_epochs | --n-epochs | Number of epochs over which to train the model. | 300 | True | None |
dropout | --dropout | Fraction of neurons affected by Dropout, from 0 to 1. | 0 | True | None |
batch_size | --batch-size | Number of time series (input and output) used in each training pass | 32 | True | None |
s_end_date | --end | The end date (format YYYY-MM-DD) to select for testing | None | True | None |
s_start_date | --start | The start date (format YYYY-MM-DD) to select for testing | None | True | None |
residuals | --residuals | Show the residuals for the model. | False | True | None |
forecast_only | --forecast-only | Do not plot the historical data without forecasts. | False | True | None |
export_pred_raw | --export-pred-raw | Export predictions to a csv file. | False | True | None |
metric | --metric | Calculate precision based on a specific metric (rmse, mse, mape) | mape | True | rmse, mse, mape, smape |
Examplesโ
2022 Jul 23, 10:36 (๐ฆ) /forecast/ $ load GME_20220719_123734.csv -a GME
2022 Jul 23, 11:01 (๐ฆ) /forecast/ $ trans GME
100%|โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ| 115/115 [00:2300:00, 4.88it/s]
Transformer model obtains MAPE: 13.11%
Actual price: $ 146.64
โโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโ
โ Datetime โ Prediction โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ 2022-07-19 00:00:00 โ $ 145.63 โ
โโโโโโโโโโโโโโโโโโโโโโโผโโโโโโโโโโโโโค
โ 2022-07-20 00:00:00 โ $ 142.28 โ
โโโโโโโโโโโโโโโโโโโโโโโผโโโโโโโโโโโโโค
โ 2022-07-21 00:00:00 โ $ 137.67 โ
โโโโโโโโโโโโโโโโโโโโโโโผโโโโโโโโโโโโโค
โ 2022-07-22 00:00:00 โ $ 137.33 โ
โโโโโโโโโโโโโโโโโโโโโโโผโโโโโโโโโโโโโค
โ 2022-07-25 00:00:00 โ $ 130.62 โ
โโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโ