๐Ÿ”ฌ DeepAries Research

CIKM 2025 Accepted

Research Overview

This demonstration showcases DeepAriesโ€”our novel reinforcement learning framework that dynamically adjusts rebalancing intervals based on market conditions. Our work presents adaptive portfolio management using deep reinforcement learning techniques for enhanced portfolio selection.

๐Ÿ”„ Adaptive Rebalancing

Dynamic rebalancing interval selection based on market conditions using reinforcement learning.

๐Ÿ’ผ Portfolio Optimization

AI-driven portfolio allocation and performance optimization with enhanced selection strategies.

โš ๏ธ Risk Management

Comprehensive risk assessment using VaR, Expected Shortfall, and stress testing methodologies.

๐Ÿค– Reinforcement Learning

Deep RL framework for financial decision making and adaptive portfolio management.

๐Ÿ“„ Related Publications

This demo accompanies our research paper:

"DeepAries: Adaptive Rebalancing Interval Selection for Enhanced Portfolio Selection"

โœ… Accepted at CIKM 2025

Authors: Jinkyu Kim, Hyunjung Yi, Mogan Gim, Donghee Choi, Jaewoo Kang

Conference: CIKM 2025 (30th ACM International Conference on Information and Knowledge Management)

Year: 2025

Keywords: Portfolio Management, Adaptive Rebalancing Interval Selection, Artificial Intelligence in Finance, Deep Reinforcement Learning

Demo Description: In this interactive demo, we showcase DeepAriesโ€”our novel reinforcement learning framework that dynamically adjusts rebalancing intervals based on market conditions. Evaluation begins on January 1, 2021, and data is continuously updated. The demo provides recommended portfolio allocation, performance comparisons (with both portfolio and benchmark index cumulative returns normalized over the selected period), adaptive rebalancing visualizations, and aggregated insights into rebalancing behavior.

๐Ÿ“ˆ Market Data

Historical market data with multiple ticker selection and default visualization

๐Ÿ“Š Index Time Series

๐Ÿ”ฎ AI Prediction Heatmap

Rows: Tickers, Columns: Dates, Values: AI Model Predictions

๐Ÿ” Technical Analysis

Advanced technical indicators and chart analysis tools to help you make informed trading decisions. Analyze price trends, momentum, and market patterns using professional-grade indicators.

๐Ÿ•ฏ๏ธ Candlestick Chart

Visualize price movements with OHLC (Open, High, Low, Close) data and volume analysis for comprehensive market insight.

๐Ÿ“ˆ Technical Indicators

Calculate and visualize various technical indicators to identify market trends and potential trading opportunities.

๐Ÿ“Š Volume Analysis

Analyze trading volume patterns to understand market sentiment and confirm price movements with volume-based indicators.

๐Ÿ”— Correlation Matrix

Analyze correlation relationships between different stocks across markets to identify diversification opportunities and market dependencies.

๐Ÿ”ฌ DeepAries Adaptive Rebalancing

Interactive demonstration of adaptive rebalancing and portfolio performance analysis

๐Ÿ’ผ Portfolio & Model Performance

Select Market, Model, and Period to analyze portfolio allocation and performance

๐Ÿ“Š Portfolio Value vs Market Index

๐Ÿงฉ Latest Allocation

Allocation on โ€”

๐Ÿ“Š Portfolio Performance

Total Return -
Sharpe Ratio -
Max Drawdown -
Volatility -
CAGR -
Sortino Ratio -
Tracking Error -
Information Ratio -
Alpha (annual) -
Beta -

๐Ÿ”„ Adaptive Rebalancing Results

Shows monthly rebalancing allocations (first trading day each month) for the selected market and model. Xโ€‘axis uses 6โ€‘month ticks.

๐Ÿ”„ Adaptive Rebalancing Timeline

Shows when rebalancing events occurred and their relationship to market volatility

๐Ÿ“Š Rebalancing Frequency Analysis

Distribution of rebalancing intervals showing adaptive behavior

๐Ÿ” Market Condition Detection

Shows market volatility, trends, and when rebalancing events occurred in relation to market conditions

โš ๏ธ Risk Management ๐Ÿšง UNDER DEVELOPMENT

Advanced risk assessment tools to evaluate portfolio risk exposure, calculate Value at Risk (VaR), and perform stress testing under various market scenarios. These tools help quantify potential losses and assess portfolio resilience under adverse market conditions.

โš ๏ธ Notice: This section is currently under development and improvement. Some features may not be fully functional or may require additional data sources.

๐Ÿ“š Key Concepts:

  • VaR (Value at Risk): Maximum potential loss over a specified time horizon with a given confidence level
  • Expected Shortfall: Average loss beyond the VaR threshold (also known as Conditional VaR)
  • Stress Testing: Simulation of extreme market conditions to assess portfolio resilience
  • Risk Metrics: Volatility, Sharpe ratio, maximum drawdown, and beta calculations

โš ๏ธ VaR (Value at Risk) Analysis

Calculate the maximum potential loss over a specified time horizon with a given confidence level using historical data analysis. VaR answers the question: "What is the worst loss I can expect over a given time period with a given confidence level?"

๐Ÿ’ก How it works: Uses historical price data to calculate daily returns, then determines the worst-case scenario based on your selected confidence level (95% or 99%). The Expected Shortfall shows the average loss beyond the VaR threshold.
โฐ Time Horizon: This represents the holding period for your investment. For example, "1 Day" means the VaR shows the maximum loss you could expect over a single trading day. "5 Days" means over a week, "30 Days" means over a month. The calculation uses daily returns and scales them to your selected time period.
VaR -
Expected Shortfall -

๐ŸŒช๏ธ Stress Testing

Simulate extreme market conditions to assess portfolio resilience and potential losses under adverse scenarios. Stress testing helps identify vulnerabilities and prepare for market crises.

๐Ÿ’ก How it works: Applies predefined shock scenarios (market crash, recession, volatility spike) to historical returns data to simulate how your portfolio would perform under extreme conditions. The chart shows the distribution of stressed returns.

๐Ÿ“Š Risk Metrics Dashboard

Comprehensive risk metrics including volatility, Sharpe ratio, and maximum drawdown for selected assets. These metrics provide a complete picture of risk-adjusted performance.

๐Ÿ’ก Key Metrics:
  • Volatility: Standard deviation of returns (higher = more risk)
  • Sharpe Ratio: Risk-adjusted return (higher = better risk-adjusted performance)
  • Max Drawdown: Largest peak-to-trough decline (lower = better)
  • Beta: Sensitivity to market movements (1.0 = market average)
Volatility -
Sharpe Ratio -
Max Drawdown -
Beta -