Kronos-Report-Java

AI Quantitative Prediction Report Generation Tool Based on Kronos Financial Large Model (Java Edition)

📖 Introduction

Kronos-Report-Java is a Java quantitative prediction tool based on the Kronos financial time-series large model, supporting stocks, funds, futures, indices and other instruments. It automatically generates multi-format analysis reports and daily briefings, and supports multiple inference backends (HTTP, ONNX Runtime, DJL).

✨ Features

Prediction & Reporting

✅ Automatic instrument type identification (stock/fund/futures/index)
✅ Automatic exchange detection
✅ Automatic filtering of futures index contracts (codes ending with 00, e.g., RB00, M00)
✅ Multiple model combinations (2k/base tokenizer + mini/small/base model)
✅ Forward/backward adjustment support
✅ Multi-format report output (TXT, HTML, PNG, PDF, CSV, charts, JSON)
✅ Rolling backtest validation (direction accuracy, MAPE)
✅ Quick backtest (validate model with recent N trading days)
✅ Backtest CSV export (record predicted vs actual OHLCV data)
✅ Retail-friendly report format

Daily Briefing

✅ Daily briefing generation (watchlist + hot stocks ranking)
✅ Watchlist priority display (watchlist items shown before hot stocks)
✅ Global model reuse (single instance, improved performance)
✅ Scheduled tasks (automatic daily briefing generation)
✅ WeChat public account promotion (QR code, subscription info, like/share buttons)

Model Inference

✅ Multi-model management (up to 3 models concurrently)
✅ Parameterized prediction (support lookback, pred_len, t, top_p, sample_count)
✅ Hot model switching (switch models without restarting service)
✅ Three inference backends: HTTP, ONNX Runtime, DJL
✅ Auto device selection (default auto, no manual specification needed)

Data Service

✅ Unified data service interface (support local files and HTTP gateway)
✅ Data source auto-switching (configure DATA_SOURCE to switch)
✅ Name mapping (stock, fund, futures, index names)
✅ Last trading day data (hot rankings, query by code)
✅ Forward adjustment calculation (based on dividend JSON files)
✅ Trading calendar (holidays, trading day judgment)
✅ Futures code case auto-correction

📦 Requirements

JDK 17+
Maven 3.6+

🔧 Build & Install

# Clone repository
git clone https://gitee.com/liebin/kronos-report.git
cd kronos-report

# Build
mvn clean package

# Install to local repository
mvn install

🚀 Usage

Single Prediction (Full Report)

# Stock
java -jar target/kronos-report-1.0.0.jar --code 000001

# Futures (automatically excludes index contracts)
java -jar target/kronos-report-1.0.0.jar --code RB2505

# Index
java -jar target/kronos-report-1.0.0.jar --code zs_000001

# Custom parameters
java -jar target/kronos-report-1.0.0.jar --code 000001 --lookback 256 --pred-len 10 --model small

# No forward adjustment
java -jar target/kronos-report-1.0.0.jar --code 000001 --no-adjusted

JSON Data Report (Pure Data Format)

# Generate JSON report (no charts or backtest)
java -jar target/kronos-report-1.0.0.jar --code 000001 --json

# Print JSON to console only, don't save file
java -jar target/kronos-report-1.0.0.jar --code 000001 --json --no-save

# With custom parameters
java -jar target/kronos-report-1.0.0.jar --code 000001 --json --lookback 256 --pred-len 10 --verbose

# JSON report without forward adjustment
java -jar target/kronos-report-1.0.0.jar --code 000001 --json --no-adjusted

Daily Briefing

# Generate briefing (quiet mode)
java -jar target/kronos-report-1.0.0.jar --briefing

# Generate briefing (verbose mode)
java -jar target/kronos-report-1.0.0.jar --briefing -v

Scheduled Task

# Start scheduler (daily at 19:30)
java -jar target/kronos-report-1.0.0.jar --scheduler

# Custom time
java -jar target/kronos-report-1.0.0.jar --scheduler --scheduler-hour 20 --scheduler-minute 0

# Stop scheduler
java -jar target/kronos-report-1.0.0.jar --scheduler-stop

📁 Project Structure

kronos-report/
├── pom.xml
├── README.md
├── README.en.md
├── LICENSE
└── src/main/java/cn/tsquant/invest/util/kronos/
    ├── KronosPredictor.java              # Main entry
    ├── config/                           # Configuration module
    │   └── KronosConfig.java
    ├── core/                             # Model inference core
    │   ├── ModelPredictor.java
    │   ├── ModelPredictorFactory.java
    │   ├── HttpModelPredictor.java
    │   ├── OnnxModelPredictor.java
    │   ├── DjlModelPredictor.java
    │   ├── PredictorWrapper.java
    │   ├── ModelConfig.java
    │   ├── ModelManager.java
    │   ├── ModelInstance.java
    │   ├── PredictRequest.java
    │   └── PredictResponse.java
    ├── data/                             # Data service module
    │   ├── DataService.java
    │   ├── DataServiceFactory.java
    │   ├── bo/                           # Business objects
    │   │   ├── AssetInfo.java
    │   │   ├── EquityItem.java
    │   │   ├── FuturesInfo.java
    │   │   ├── FuturesParsed.java
    │   │   ├── KLine.java
    │   │   └── TradeDates.java
    │   ├── gateway/                      # HTTP gateway implementation
    │   │   └── GatewayDataService.java
    │   └── local/                        # Local file implementation
    │       ├── LocalDataService.java
    │       ├── LastDayDataLoader.java
    │       ├── Mappings.java
    │       └── ForwardAdjusted.java
    ├── report/                           # Report generation module
    │   ├── Backtest.java
    │   ├── ChartGenerator.java
    │   ├── TxtReport.java
    │   ├── HtmlReport.java
    │   ├── ImageReport.java
    │   └── JsonReport.java
    ├── briefing/                         # Daily briefing module
    │   ├── BriefingConfig.java
    │   ├── BriefingGenerator.java
    │   ├── HtmlBriefing.java
    │   └── TxtBriefing.java
    └── utils/                            # Utilities
        ├── TradeDateUtils.java
        ├── TradeCalendar.java
        ├── CodeUtils.java
        ├── CsvLoader.java
        ├── AdjustedUtils.java
        ├── FileUtils.java
        └── TypeUtils.java

📝 Command Line Parameters

Parameter	Default	Description
`--code, -c`	Required	Instrument code
`--start-date, -s`	None	Start date YYYYMMDD
`--end-date, -e`	None	End date YYYYMMDD
`--pred-len, -p`	5	Prediction steps
`--lookback, -l`	512	Lookback window
`--t`	0.6	Temperature parameter
`--top-p`	0.8	Top-p sampling parameter
`--sample-count`	5	Sample count
`--tokenizer`	2k	Tokenizer type
`--model, -m`	mini	Model type
`--device, -d`	auto	Device type
`--no-save`	false	Don't save files
`--adjusted`	true	Forward adjustment (enabled by default)
`--no-adjusted`	false	No forward adjustment
`--json`	false	Generate JSON data report
`--briefing`	false	Generate daily briefing
`--verbose, -v`	false	Verbose output
`--scheduler`	false	Start scheduler
`--scheduler-stop`	false	Stop scheduler

📂 Output Directory Structure

/home/liebin/dev/data/report/
├── AI预测/                           # Prediction reports
│   └── {date}/{name}/{tokenizer}_{model}_{lookback}/
│       ├── AI预测报告_{name}_{date}.html
│       ├── AI预测报告_{name}_{date}.txt
│       ├── AI预测报告_{name}_{date}.png
│       ├── AI预测报告_{name}_{date}.pdf
│       ├── AI预测报告_{name}_{date}.json
│       ├── {name}_chart_{date}.png
│       ├── {name}_trend_{date}.png
│       ├── {name}_backtest_{date}.png
│       ├── {name}_backtest_data_{date}.csv
│       └── {name}_predictions_{date}.csv
└── AI简报/                           # Daily briefings
    └── {date}/
        ├── AI量化简报_{date}.html
        ├── AI量化简报_{date}.txt
        └── AI量化简报_{date}.png

📊 JSON Report Format

{
  "code": "000001",
  "name": "Ping An Bank",
  "timestamp": "2026-06-11T15:30:00",
  "last_trade_date": "2026-06-10",
  "model": "2k_mini",
  "lookback": 512,
  "pred_len": 5,
  "adjusted": true,
  "summary": {
    "current_price": 11.20,
    "predicted_price": 11.35,
    "change_pct": 1.34,
    "trend": "Bullish 📈",
    "low": 11.10,
    "high": 11.50
  },
  "recent": [
    {
      "d": "2026-06-04",
      "o": 11.05,
      "h": 11.15,
      "l": 11.00,
      "c": 11.10,
      "v": 1000000,
      "a": 11100000
    }
  ],
  "predicted": [
    {
      "d": "2026-06-11",
      "o": 11.22,
      "h": 11.38,
      "l": 11.15,
      "c": 11.30,
      "v": 1020000,
      "a": 11500000
    }
  ]
}

Field Description

Field	Type	Description
`code`	string	Instrument code
`name`	string	Instrument name
`timestamp`	string	Report generation time
`last_trade_date`	string	Last trading date
`model`	string	Model identifier used
`lookback`	int	Lookback window length
`pred_len`	int	Prediction steps
`adjusted`	boolean	Whether forward adjustment is applied
`summary`	object	Prediction summary
`recent`	array	Last 30 trading days K-lines
`predicted`	array	Next N trading days predicted K-lines

K-line Field Abbreviation Mapping

Abbr	Full Name	Description
`d`	date	Date
`o`	open	Open price
`h`	high	High price
`l`	low	Low price
`c`	close	Close price
`v`	volume	Volume
`a`	amount	Amount

🔌 Inference Backends

Backend	Description	Priority
HTTP	Remote Python model service	1
ONNX Runtime	Local ONNX model inference	2
DJL	Local PyTorch model inference	3

⚙️ Configuration

Edit KronosConfig.java to modify settings:

// Data source configuration
public static final String DATA_SOURCE = "local";  // local / http

// Model inference configuration
public static final boolean USE_HTTP_FIRST = true;
public static final String MODEL_HTTP_HOST = "http://localhost";
public static final int MODEL_HTTP_PORT = 16888;

// Model management
public static final int MAX_MODELS = 3;

// Default model
public static final String DEFAULT_TOKENIZER = "2k";
public static final String DEFAULT_MODEL = "mini";

📄 License

This project is licensed under the AGPL-3.0 License.

💬 Technical Support & Services

If you encounter any issues or need technical services, feel free to contact the author:

🤖 AI Customer Service Development: Intelligent Q&A, auto-reply, multi-channel integration
🧠 LLM Application Development: Model fine-tuning, Prompt engineering, RAG systems
💻 Software Development: Java&Python backend, quantitative systems, data processing
📈 Quantitative Strategy Development: Factor mining, backtesting systems, live trading integration
🔧 Custom Development: Tailored AI quantitative solutions

Scan WeChat to add the author, please state your purpose

⚠️ Disclaimer

Prediction results are generated by AI models and are for reference only. They do not constitute investment advice.

FilesExpand file tree

README.en.md

Latest commit

History

README.en.md

File metadata and controls

Kronos-Report-Java

📖 Introduction

✨ Features

Prediction & Reporting

Daily Briefing

Model Inference

Data Service

📦 Requirements

🔧 Build & Install

🚀 Usage

Single Prediction (Full Report)

JSON Data Report (Pure Data Format)

Daily Briefing

Scheduled Task

📁 Project Structure

📝 Command Line Parameters

📂 Output Directory Structure

📊 JSON Report Format

Field Description

K-line Field Abbreviation Mapping

🔌 Inference Backends

⚙️ Configuration

📄 License

💬 Technical Support & Services

⚠️ Disclaimer