Stock Prediction demo setup

Pre-requisites:

Pivotal GemFire 8.1+ installed
Spring XD 1.1+ installed
R 3.1+ installed with the following packages
1. RCurl
2. quantmod
3. TTR
4. jsonlite
5. RSNNS

Installation

Install R:

Linux

    su -c 'rpm -Uvh http://download.fedoraproject.org/pub/epel/6/i386/epel-release-6-8.noarch.rpm'
    sudo yum update
    sudo yum install R

Mac OSX (homebrew)

 brew install R

The R packages can be easily installed by typing at the R command line:

> install.packages(c("RCurl","quantmod","TTR","jsonlite","RSNNS"))

Pivotal GemFire Setup

Execute the setup.gfsh script as follows:

gfsh run --file=setup.gfsh

1. Executing - start locator --name=locator1 --J=-Dgemfire.http-service-port=7575

.............................
Locator in FinanceStream/locator1 on 192.168.3.5[10334] as locator1 is currently online.
Process ID: 34681
Uptime: 15 seconds
...
Successfully connected to: [host=192.168.3.5, port=1099]

Cluster configuration service is up and running.

2. Executing - start server --name=server1 --J=-Dgemfire.start-dev-rest-api=true --J=-Dgemfire.http-service-port=8080

..........
Server in FinanceStream/server1 on 192.168.3.5[40404] as server1 is currently online.
Process ID: 34683
Uptime: 5 seconds
...

3. Executing - create region --name=/Stocks --type=PARTITION

Member  | Status
------- | -------------------------------------
server1 | Region "/Stocks" created on "server1"

4. Executing - import data --file=../Stocks.gfd --region=/Stocks --member=server1

Data imported from file : FinanceStream/Stocks.gfd on host : 192.168.3.5 to region : /Stocks

5. Executing - describe --region=/Stocks

Name            : Stocks
Data Policy     : partition
Hosting Members : server1

Non-Default Attributes Shared By Hosting Members

 Type  | Name | Value
------ | ---- | -----
Region | size | 2601

Test the server REST endpoint

In a web browser, access http://localhost:8080/gemfire-api/v1

{
  "regions" : [ {
    "name" : "Stocks",
    "type" : "REPLICATE",
    "key-constraint" : null,
    "value-constraint" : null
  } ]
}

Spring XD

Create streams in Spring XD

Start with the base stream

xd:> script --file stream-base.xd

stream create process_sink --definition "queue:stocks >  transform --script='file:transform.groovy' | gemfire-json-server --useLocator=true --host=localhost --port=10334 --regionName=Stocks --keyExpression=payload.getField('entryTimestamp')" --deploy
Created and deployed new stream 'process_sink'
stream create http_source --definition "http --port=9000 | splitter --expression=#jsonPath(payload,'$') > queue:stocks" --deploy
Created and deployed new stream 'http_source'

If you want to show online quotes (need to be in market hours):

xd:> script --file stream-online-quotes.xd

stream create yahoo_finance_source --definition "trigger --fixedDelay=1 | http-client --url='''https://query.yahooapis.com/v1/public/yql?q=select * from yahoo.finance.quote where symbol in (\"TSLA\")&format=json&env=store://datatables.org/alltableswithkeys''' --httpMethod=GET | splitter --expression=#jsonPath(payload,'$.query.results.quote')  > queue:stocks" --deploy
Created and deployed new stream 'yahoo_finance_source'

If you want to show predictions:

xd:> script --file stream-predictions.xd

stream create r_process --definition "tap:stream:process_sink.transform > object-to-json | shell --command='Rscript nn_evaluate_jordan_ema.R' | splitter --expression=#jsonPath(payload,'$') | gemfire-json-server --useLocator=true --host=localhost --port=10334 --regionName=Predictions --keyExpression=payload.getField('entryTimestamp') " --deploy
Created and deployed new stream 'r_process'
stream create trainstream --definition "trigger --fixedDelay=10 | shell --command='Rscript nn_train_jordan_ema.R' | log " --deploy
Created and deployed new stream 'trainstream'

Using the simulator:

The simulator replays random 1000 quotes previously stored on GemFire / Geode

$ cd FinanceStreamSimulator
$ ./gradlew run

UI using JavaFX

The UI for the application uses JavaFX and acting as GemFire client can subscribe to event notifications on the regions and receive updates as soon as data is updated on the servers, leveraging a PUSH instead of PULL strategy, which is much more efficient.

Under JavaFXChart directory you can build/execute the client:

$ cd JavaFXChart
$ ./gradlew run

ganger85/StockPrediction