[Alpha] Add new alpha and alpha evaluation methods

[CYX22222003]

This pull request introduces an autoregressive (AR) model for time series prediction and adds an evaluation framework for alpha signals, along with corresponding tests. The main themes are the implementation of the AR model, integration into the alpha signal pipeline, and the addition of an information coefficient (IC) evaluation utility.

Key changes:

Autoregressive Model Implementation

• Added AR model class and related utilities for training, inference, and dataset preparation in autoregression.py, including functions to build training data, train the model, and load a pretrained model.
• Introduced a data processing module (data_processing.py) to load financial time series data and compute log returns.

Integration with Alpha Pipeline

• Implemented AutoregressionAlpha, an IAlpha subclass, which computes AR-based predictions and adds them as a new column (ar_pred) to the DataFrame.

Alpha Signal Evaluation

• Added AlphaEvaluator class to interface.py, which computes the information coefficient (Spearman correlation) between specified alpha signals and future returns over multiple timeframes.

Testing

• Added tests for AutoregressionAlpha to verify the presence and correctness of the ar_pred column and its handling of short series.
• Added tests for AlphaEvaluator to ensure correct output format, value types, and expected behavior on constant alpha signals.

[Copilot]

Pull request overview

Adds an autoregressive (AR) alpha implementation and a new alpha evaluation utility (information coefficient) with corresponding unit tests, extending the alpha pipeline with model-based signals and evaluation tooling.

Changes:

• Introduces AR model utilities + a bundled pretrained weights file under alpha/AutoRegression/.
• Adds AutoregressionAlpha that computes an ar_pred signal column from AR predictions.
• Adds AlphaEvaluator to compute Spearman IC vs. forward returns, plus new tests for AR alpha and IC evaluation.

Reviewed changes

Copilot reviewed 6 out of 7 changed files in this pull request and generated 14 comments.

Show a summary per file

File	Description
tests/test_alpha.py	Adds tests for AutoregressionAlpha and AlphaEvaluator.
alpha/interface.py	Adds AlphaEvaluator and forward-return timeframe configuration.
alpha/AutoregressionAlpha.py	Implements AR-based alpha signal (ar_pred) using a pretrained torch model.
alpha/AutoRegression/data_processing.py	Adds utilities to load data and compute log returns (training/experimentation support).
alpha/AutoRegression/autoregression.py	Adds AR model definition, training/inference helpers, and model-loading logic.
alpha/AutoRegression/ar_model.pth	Adds pretrained AR model weights artifact.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.