# 230723pp第六组

**Repository Path**: APIJSON/real-time-fraud-risk

## Basic Information

- **Project Name**: 230723pp第六组
- **Description**: 基于 Kafka、Spark Streaming、MySQL、APIJSON 和 ECharts 构建的实时金融交易欺诈风险平台。本仓库为副本，请给原仓库右上角点亮 ⭐️ Star
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: main
- **Homepage**: https://gitee.com/yi-huaimin/real-time-fraud-risk
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 1
- **Created**: 2026-08-09
- **Last Updated**: 2026-08-09

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# Real-time Fraud Risk Platform

Course-cluster implementation of real-time financial fraud detection.

## Verified architecture

| Component | Deployment |
|---|---|
| Kafka 2.5.0 | Three brokers on `niit-master`, `niit-slave1`, and `niit-slave2`, all on port 9092 |
| Topic | `bank_transaction`, 3 partitions, replication factor 3, `min.insync.replicas=2` |
| Spark Streaming 2.4.6 | One Direct-mode application on `spark://niit-master:7077`, Scala 2.11, 3-second batches |
| HDFS | `hdfs://niit-master:9000/checkpoint/fraud` and `/warehouse/realrisk` |
| Hive 2.3.7 | Five external ORC tables with Snappy compression, queried through HiveServer2/Beeline |
| MySQL 5.7.16 | Transaction detail, accumulated channel statistics, alerts, window statistics, and blacklist |
| API | Spring Boot and APIJSON 8.2 on port 18080 |
| Dashboard | ECharts 5.6, overview, trend, channels, heatmap, R01-R04, regions, and alerts |

The Streaming artifact targets Java 8, Scala 2.11, and Spark 2.4.6. The API is built separately because it uses a newer Spring runtime.

## Cluster nodes

| Host | IP | Roles |
|---|---|---|
| `niit-master` | `192.168.10.102` | HDFS NameNode, Spark Master, Kafka broker, MySQL, HiveServer2, API |
| `niit-slave1` | `192.168.10.103` | HDFS DataNode, Spark Worker, Kafka broker, YARN ResourceManager/NodeManager |
| `niit-slave2` | `192.168.10.104` | HDFS DataNode, Spark Worker, Kafka broker, YARN NodeManager |

## Build

```powershell
$env:JAVA_HOME='C:\Program Files\Java\jdk1.8.0_144'
mvn -f streaming-pom.xml -DskipTests clean package

$env:JAVA_HOME='C:\Program Files\Java\jdk-21'
mvn -f api-pom.xml -DskipTests clean package
```

## Deployment

The VM script expects passwordless SSH and pre-created root-only credential files on the master:

- `/opt/realrisk/conf/app.env` (`0600`)
- `/opt/realrisk/conf/mysql-admin.cnf` (`0600`)

It never stores database passwords in Git. It preserves shared ZooKeeper, Hadoop, Hive, and Spark services.

```powershell
powershell -ExecutionPolicy Bypass -File scripts\deploy-vm.ps1 -TransactionCount 5000
```

The formal Producer always uses a random 50-500 ms delay. Do not use the deployment script to resume a completed formal run because it intentionally recreates project data. Post-processing recovery procedures are in [docs/CLUSTER_RUNBOOK.md](docs/CLUSTER_RUNBOOK.md).

Dashboard: `http://192.168.10.102:18080/`

Stop only project-owned API, Streaming, and Kafka processes:

```powershell
powershell -ExecutionPolicy Bypass -File scripts\stop-vm.ps1
```

The stop script does not stop or power off shared cluster services.

## Formal result

The 2026-07-28 formal run produced 5,000 successful messages with zero failures in 1,463,537 ms. MySQL contains 5,000 transactions, channel totals of 5,000, and 68 alerts. See [docs/VM_CLUSTER_ACCEPTANCE.md](docs/VM_CLUSTER_ACCEPTANCE.md) and `evidence/vm-cluster/`.
# Vue 3 Dashboard

The production dashboard is implemented in `frontend/` with Vue 3, Vite,
TypeScript, Vue Router, Pinia, Axios, and ECharts 5.6. All browser assets are
bundled locally and remain usable when the VM has no Internet access.

```powershell
cd frontend
npm ci
npm run type-check
npm run dev       # http://localhost:5173, /api proxies to the master VM
npm run build
```

IDEA shared configurations include `Vue Dev`, `Vue Build`,
`Financial Risk-Vue Development`, and `Financial Risk-Deploy Vue Page`. The
deployment configuration builds the frontend and API, backs up the remote JAR,
restarts only the project API, validates the formal `5000 / 5000 / 68` counts,
and restores the backup automatically if validation fails. The previous page
is retained at `/legacy.html`.

## IDEA green run

Open `C:\Users\86132\fraud-risk-deployment-20260727` in IntelliJ IDEA and select
`Financial Risk - Start and Open Dashboard` (the shared Chinese-named
Application configuration). Press the green Run button. The Java launcher
invokes only `scripts/start-existing-vm.ps1`, which powers on missing VMs,
restores missing services, validates the formal `5000 / 5000 / 68` data, and
opens `http://192.168.10.102:18080/`.

The `Financial Risk - Stop Project Application Only` configuration passes
`--stop` and stops only realrisk-owned API/legacy Streaming PIDs. Full formal
redeployment is available solely through an explicit command-line `--deploy`;
no shared IDEA configuration uses it.

For frontend and backend development together, select the shared Application
configuration `Financial Risk - Frontend/Backend One-click Start` (Chinese name
in IDEA). It restores the existing remote backend, starts Vite on port 5173,
and opens a new browser window. IDEA's Stop button terminates only the local
Vite development process; shared VM services and formal data remain intact.