The Temporal Fusion Transformers (TFT) is a sophisticated mannequin for time collection forecasting. It consists of the Variable Choice Community (VSN), which is a key part of the mannequin. It’s particularly designed to mechanically establish and concentrate on probably the most related options inside a dataset. It achieves this by assigning realized weights to every enter variable, successfully highlighting which options contribute most to the predictive process.
This VSN-based method will probably be our second discount approach. We’ll implement it utilizing PyTorch Forecasting, which permits us to leverage the Variable Choice Community from the TFT mannequin.
We’ll use a fundamental configuration. Our aim isn’t to create the highest-performing mannequin potential, however moderately to establish probably the most related options whereas utilizing minimal assets.
from pytorch_forecasting import TemporalFusionTransformer, TimeSeriesDataSet
from pytorch_forecasting.metrics import QuantileLoss
from lightning.pytorch.callbacks import EarlyStopping
import lightning.pytorch as pl
import torchpl.seed_everything(42)
max_encoder_length = 32
max_prediction_length = 1
VAL_SIZE = .2
VARIABLES_IMPORTANCE = .8
model_data_feature_sel = initial_model_train.be a part of(stationary_df_train)
model_data_feature_sel = model_data_feature_sel.be a part of(pca_df_train)
model_data_feature_sel['price'] = model_data_feature_sel['price'].astype(float)
model_data_feature_sel['y'] = model_data_feature_sel['price'].pct_change()
model_data_feature_sel = model_data_feature_sel.iloc[1:].reset_index(drop=True)
model_data_feature_sel['group'] = 'spy'
model_data_feature_sel['time_idx'] = vary(len(model_data_feature_sel))
train_size_vsn = int((1-VAL_SIZE)*len(model_data_feature_sel))
train_data_feature = model_data_feature_sel[:train_size_vsn]
val_data_feature = model_data_feature_sel[train_size_vsn:]
unknown_reals_origin = [col for col in model_data_feature_sel.columns if col.startswith('value_')] + ['y']
timeseries_config = {
"time_idx": "time_idx",
"goal": "y",
"group_ids": ["group"],
"max_encoder_length": max_encoder_length,
"max_prediction_length": max_prediction_length,
"time_varying_unknown_reals": unknown_reals_origin,
"add_relative_time_idx": True,
"add_target_scales": True,
"add_encoder_length": True
}
training_ts = TimeSeriesDataSet(
train_data_feature,
**timeseries_config
)
The VARIABLES_IMPORTANCE threshold is ready to 0.8, which suggests we’ll retain options within the high eightieth percentile of significance as decided by the Variable Choice Community (VSN). For extra details about the Temporal Fusion Transformers (TFT) and its parameters, please discuss with the documentation.
Subsequent, we’ll prepare the TFT mannequin.
if torch.cuda.is_available():
accelerator = 'gpu'
num_workers = 2
else :
accelerator = 'auto'
num_workers = 0validation = TimeSeriesDataSet.from_dataset(training_ts, val_data_feature, predict=True, stop_randomization=True)
train_dataloader = training_ts.to_dataloader(prepare=True, batch_size=64, num_workers=num_workers)
val_dataloader = validation.to_dataloader(prepare=False, batch_size=64*5, num_workers=num_workers)
tft = TemporalFusionTransformer.from_dataset(
training_ts,
learning_rate=0.03,
hidden_size=16,
attention_head_size=2,
dropout=0.1,
loss=QuantileLoss()
)
early_stop_callback = EarlyStopping(monitor="val_loss", min_delta=1e-5, endurance=5, verbose=False, mode="min")
coach = pl.Coach(max_epochs=20, accelerator=accelerator, gradient_clip_val=.5, callbacks=[early_stop_callback])
coach.match(
tft,
train_dataloaders=train_dataloader,
val_dataloaders=val_dataloader
)
We deliberately set max_epochs=20 so the mannequin doesn’t prepare too lengthy. Moreover, we carried out an early_stop_callback that halts coaching if the mannequin exhibits no enchancment for five consecutive epochs (endurance=5).
Lastly, utilizing the most effective mannequin obtained, we choose the eightieth percentile of crucial options as decided by the VSN.
best_model_path = coach.checkpoint_callback.best_model_path
best_tft = TemporalFusionTransformer.load_from_checkpoint(best_model_path)raw_predictions = best_tft.predict(val_dataloader, mode="uncooked", return_x=True)
def get_top_encoder_variables(best_tft,interpretation):
encoder_importances = interpretation["encoder_variables"]
sorted_importances, indices = torch.type(encoder_importances, descending=True)
cumulative_importances = torch.cumsum(sorted_importances, dim=0)
threshold_index = torch.the place(cumulative_importances > VARIABLES_IMPORTANCE)[0][0]
top_variables = [best_tft.encoder_variables[i] for i in indices[:threshold_index+1]]
if 'relative_time_idx' in top_variables:
top_variables.take away('relative_time_idx')
return top_variables
interpretation= best_tft.interpret_output(raw_predictions.output, discount="sum")
top_encoder_vars = get_top_encoder_variables(best_tft,interpretation)
print(f"nOriginal variety of options: {stationary_df_train.form[1]}")
print(f"Variety of options after Variable Choice Community (VSN): {len(top_encoder_vars)}n")
The unique dataset contained 438 options, which had been then lowered to 1 characteristic solely after making use of the VSN technique! This drastic discount suggests a number of prospects:
- Most of the authentic options might have been redundant.
- The characteristic choice course of might have oversimplified the info.
- Utilizing solely the goal variable’s historic values (autoregressive method) may carry out in addition to, or presumably higher than, fashions incorporating exogenous variables.
On this ultimate part, we evaluate out discount strategies utilized to our mannequin. Every technique is examined whereas sustaining similar mannequin configurations, various solely the options subjected to discount.
We’ll use TiDE, a small state-of-the-art Transformer-based mannequin. We’ll use the implementation offered by NeuralForecast. Any mannequin from NeuralForecast right here would work so long as it permits exogenous historic variables.
We’ll prepare and check two fashions utilizing each day SPY (S&P 500 ETF) knowledge. Each fashions may have the identical:
- Prepare-test cut up ratio
- Hyperparameters
- Single time collection (SPY)
- Forecasting horizon of 1 step forward
The one distinction between the fashions would be the characteristic discount approach. That’s it!
- First mannequin: Unique options (no characteristic discount)
- Second mannequin: Function discount utilizing PCA
- Third mannequin: Function discount utilizing VSN
This setup permits us to isolate the impression of every characteristic discount approach on mannequin efficiency.
First we prepare the three fashions with the identical configuration aside from the options.
from neuralforecast.fashions import TiDE
from neuralforecast import NeuralForecasttrain_data = initial_model_train.be a part of(stationary_df_train)
train_data = train_data.be a part of(pca_df_train)
test_data = initial_model_test.be a part of(stationary_df_test)
test_data = test_data.be a part of(pca_df_test)
hist_exog_list_origin = [col for col in train_data.columns if col.startswith('value_')] + ['y']
hist_exog_list_pca = [col for col in train_data.columns if col.startswith('PC')] + ['y']
hist_exog_list_vsn = top_encoder_vars
tide_params = {
"h": 1,
"input_size": 32,
"scaler_type": "sturdy",
"max_steps": 500,
"val_check_steps": 20,
"early_stop_patience_steps": 5
}
model_original = TiDE(
**tide_params,
hist_exog_list=hist_exog_list_origin,
)
model_pca = TiDE(
**tide_params,
hist_exog_list=hist_exog_list_pca,
)
model_vsn = TiDE(
**tide_params,
hist_exog_list=hist_exog_list_vsn,
)
nf = NeuralForecast(
fashions=[model_original, model_pca, model_vsn],
freq='D'
)
val_size = int(train_size*VAL_SIZE)
nf.match(df=train_data,val_size=val_size,use_init_models=True)
Then, we make the predictions.
from tabulate import tabulate
y_hat_test_ret = pd.DataFrame()
current_train_data = train_data.copy()y_hat_ret = nf.predict(current_train_data)
y_hat_test_ret = pd.concat([y_hat_test_ret, y_hat_ret.iloc[[-1]]])
for i in vary(len(test_data) - 1):
combined_data = pd.concat([current_train_data, test_data.iloc[[i]]])
y_hat_ret = nf.predict(combined_data)
y_hat_test_ret = pd.concat([y_hat_test_ret, y_hat_ret.iloc[[-1]]])
current_train_data = combined_data
predicted_returns_original = y_hat_test_ret['TiDE'].values
predicted_returns_pca = y_hat_test_ret['TiDE1'].values
predicted_returns_vsn = y_hat_test_ret['TiDE2'].values
predicted_prices_original = []
predicted_prices_pca = []
predicted_prices_vsn = []
for i in vary(len(predicted_returns_pca)):
if i == 0:
last_true_price = train_data['price'].iloc[-1]
else:
last_true_price = test_data['price'].iloc[i-1]
predicted_prices_original.append(last_true_price * (1 + predicted_returns_original[i]))
predicted_prices_pca.append(last_true_price * (1 + predicted_returns_pca[i]))
predicted_prices_vsn.append(last_true_price * (1 + predicted_returns_vsn[i]))
true_values = test_data['price']
strategies = ['Original','PCA', 'VSN']
predicted_prices = [predicted_prices_original,predicted_prices_pca, predicted_prices_vsn]
outcomes = []
for technique, costs in zip(strategies, predicted_prices):
mse = np.imply((np.array(costs) - true_values)**2)
rmse = np.sqrt(mse)
mae = np.imply(np.abs(np.array(costs) - true_values))
outcomes.append([method, mse, rmse, mae])
headers = ["Method", "MSE", "RMSE", "MAE"]
desk = tabulate(outcomes, headers=headers, floatfmt=".4f", tablefmt="grid")
print("nPrediction Errors Comparability:")
print(desk)
with open("prediction_errors_comparison.txt", "w") as f:
f.write("Prediction Errors Comparability:n")
f.write(desk)
We forecast the each day returns utilizing the mannequin, then convert these again to costs. This method permits us to calculate prediction errors utilizing costs and evaluate the precise costs to the forecasted costs in a plot.
The same efficiency of the TiDE mannequin throughout each authentic and lowered characteristic units reveals a vital perception: characteristic discount didn’t result in improved predictions as one may anticipate. This implies potential key points:
- Info loss: regardless of aiming to protect important knowledge, dimensionality discount strategies discarded info related to the prediction process, explaining the shortage of enchancment with fewer options.
- Generalization struggles: constant efficiency throughout characteristic units signifies the mannequin’s issue in capturing underlying patterns, no matter characteristic depend.
- Complexity overkill: related outcomes with fewer options recommend TiDE’s refined structure could also be unnecessarily advanced. A less complicated mannequin, like ARIMA, might probably carry out simply as effectively.
Then, let’s study the chart to see if we will observe any vital variations among the many three forecasting strategies and the precise costs.
import matplotlib.pyplot as pltplt.determine(figsize=(12, 6))
plt.plot(train_data['ds'], train_data['price'], label='Coaching Information', shade='blue')
plt.plot(test_data['ds'], true_values, label='True Costs', shade='inexperienced')
plt.plot(test_data['ds'], predicted_prices_original, label='Predicted Costs', shade='crimson')
plt.legend()
plt.title('SPY Worth Forecast Utilizing All Unique Function')
plt.xlabel('Date')
plt.ylabel('SPY Worth')
plt.savefig('spy_forecast_chart_original.png', dpi=300, bbox_inches='tight')
plt.shut()
plt.determine(figsize=(12, 6))
plt.plot(train_data['ds'], train_data['price'], label='Coaching Information', shade='blue')
plt.plot(test_data['ds'], true_values, label='True Costs', shade='inexperienced')
plt.plot(test_data['ds'], predicted_prices_pca, label='Predicted Costs', shade='crimson')
plt.legend()
plt.title('SPY Worth Forecast Utilizing PCA Dimensionality Discount')
plt.xlabel('Date')
plt.ylabel('SPY Worth')
plt.savefig('spy_forecast_chart_pca.png', dpi=300, bbox_inches='tight')
plt.shut()
plt.determine(figsize=(12, 6))
plt.plot(train_data['ds'], train_data['price'], label='Coaching Information', shade='blue')
plt.plot(test_data['ds'], true_values, label='True Costs', shade='inexperienced')
plt.plot(test_data['ds'], predicted_prices_vsn, label='Predicted Costs', shade='crimson')
plt.legend()
plt.title('SPY Worth Forecast Utilizing VSN')
plt.xlabel('Date')
plt.ylabel('SPY Worth')
plt.savefig('spy_forecast_chart_vsn.png', dpi=300, bbox_inches='tight')
plt.shut()
The distinction between true and predicted costs seems constant throughout all three fashions, with no noticeable variation in efficiency between them.
We did it! We explored the significance of characteristic discount in time collection evaluation and offered a sensible implementation information:
- Function discount goals to simplify fashions whereas sustaining predictive energy. Advantages embrace lowered complexity, improved generalization, simpler interpretation, and computational effectivity.
- We demonstrated two discount strategies utilizing FRED knowledge:
- Principal Element Evaluation (PCA), a linear dimensionality discount technique, lowered options from 438 to 76 whereas retaining 90% of defined variance.
- Variable Choice Community (VSN) from the Temporal Fusion Transformers, a non-linear method, drastically lowered options to simply 1 utilizing an eightieth percentile significance threshold.
- Analysis utilizing TiDE fashions confirmed related efficiency throughout authentic and lowered characteristic units, suggesting characteristic discount might not at all times enhance forecasting efficiency. This might be as a result of info loss throughout discount, the mannequin’s issue in capturing underlying patterns, or the chance {that a} less complicated mannequin is perhaps equally efficient for this explicit forecasting process.
On a ultimate be aware, we didn’t discover all characteristic discount strategies, corresponding to SHAP (SHapley Additive exPlanations), which supplies a unified measure of characteristic significance throughout numerous mannequin varieties. Even when we didn’t enhance our mannequin, it’s nonetheless higher to carry out characteristic curation and evaluate efficiency throughout totally different discount strategies. This method helps make sure you’re not discarding precious info whereas optimizing your mannequin’s effectivity and interpretability.
In future articles, we’ll apply these characteristic discount strategies to extra advanced fashions, evaluating their impression on efficiency and interpretability. Keep tuned!
Able to put these ideas into motion? You’ll find the entire code implementation right here.
👏 Clap it as much as 50 instances
🤝 Ship me a LinkedIn connection request to remain in contact
Your help means every little thing! 🙏