skills.rest
aliyun
aliyunOFFICIAL
Explore all skills by this author →
WebsiteGitHub

alternative-factor-mining

Mines alpha factors from alternative data and validates them with IC and group-return tests.

606|68|Updated Apr 21, 2025
Search Tags:#sentiment analysis#backtesting#quantitative research#alpha#alternative data#factor mining#ic analysis
One-click install
npx skills add https://github.com/aliyun/qwen-dianjin --skill alternative-factor-mining
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: alternative-factor-mining
Source: https://github.com/aliyun/qwen-dianjin/tree/main/DianJin-SKILLS/investment-researcher/alternative-factor-mining
Command: npx skills add https://github.com/aliyun/qwen-dianjin --skill alternative-factor-mining

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires pandas, numpy, scipy.

What problem does it solve?

Quantitative researchers struggle to systematically discover and validate excess-return sources beyond traditional price-volume and financial factors. This Skill provides an end-to-end workflow for turning alternative data (news sentiment, institutional research records, announcements, investor Q&A) into tested, economically grounded factors.

Core Features & Use Cases

  • Alternative Data Acquisition: Retrieves sentiment news, institutional investigation records, A-share announcements, and investor interaction Q&A via the gildata-aidata service, supplemented by web search for social media sentiment.
  • Factor Construction: Builds quantifiable factors such as sentiment, attention, divergence, management confidence, and expectation-surprise factors with documented economic rationale.
  • Rigorous Validation: Runs Rank IC analysis, factor correlation checks, and grouped return tests in Python, then decomposes excess returns by industry, market cap, style exposure, and market regime.
  • Use Case: A financial engineering researcher wants to test whether institutional investigation records predict stock returns. The Skill fetches the records, constructs a management-confidence factor, computes IC/ICIR and quintile returns, and produces a standardized Markdown report with investment recommendations.

Quick Start

Ask the agent to mine alternative-data factors for a stock such as CATL using sentiment news and institutional investigation records, then validate the factors with IC analysis and produce a full research report.

Frequently Asked Questions about alternative-factor-mining

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I build alpha factors from alternative data like news sentiment?▼

Fetch sentiment news, investigation records, and announcements through the gildata-aidata service, then convert them into quantifiable factors such as sentiment scores, attention counts, and divergence measures. Validate each factor with Rank IC analysis and grouped return tests before use.

How to test if a stock factor is effective using IC analysis?▼

Compute the Spearman rank correlation between factor values and future N-period returns to get the Rank IC. A factor is considered effective when mean IC exceeds 0.03, ICIR exceeds 0.5, and the proportion of positive IC periods is above 55 percent.

What alternative data sources can be used for stock factor mining?▼

The workflow uses stock and industry news sentiment, institutional investigation records, A-share announcements, and investor interaction platform Q&A retrieved via gildata-aidata. Social media discussion heat from platforms like stock forums can be added through web search.

How do I check if a new factor is redundant with existing factors?▼

Compute a correlation matrix between the new factor and existing factors such as market cap, valuation, momentum, and volatility. Factor pairs with absolute correlation above 0.7 are flagged as highly redundant and should be deduplicated.

What are the limitations of alternative data factors?▼

Alternative factors are prone to overfitting and data-mining bias, often have short effective windows due to fast information decay, and may involve high turnover and transaction costs. Every factor needs a sound economic rationale and ideally out-of-sample or rolling-window validation.

Related Skills

Explore More Finance & Accounting Skills→
AshesOfTheUndeadAshesOfTheUndead

finance

Retrieve real-time market quotes, historical prices, and financial statements.

Community
Advanced
copyleftdevcopyleftdev

aqr-factor-investing

Construct value, momentum, and quality factors with risk-managed portfolio backtesting.

Community
Advanced
GeeksfinoGeeksfino
271

sector-rotation-detector

Analyze macro indicators to generate sector rotation signals for 31 Shenwan industries.

Community
Advanced
Tommy-ywTommy-yw
539

sparse-autoencoder-training

Train and analyze Sparse Autoencoders to decompose neural network activations into interpretable features.

Community
Advanced