431 lines
17 KiB
JavaScript
431 lines
17 KiB
JavaScript
exports.newMachineLearningBotModulesLearningSystem = function (processIndex) {
|
|
/*
|
|
The Learning System is the user defined set of rules compliant with the Learning Protocol that
|
|
defines the learning logic to be applied during each cycle of the Simulation.
|
|
*/
|
|
const MODULE_NAME = 'Learning System'
|
|
let thisObject = {
|
|
loadModel: loadModel,
|
|
saveModel: saveModel,
|
|
mantain: mantain,
|
|
reset: reset,
|
|
run: run,
|
|
updateChart: updateChart,
|
|
initialize: initialize,
|
|
finalize: finalize
|
|
}
|
|
|
|
/*
|
|
These 3 are the main data structures available to users
|
|
when writing conditions and formulas.
|
|
*/
|
|
let chart
|
|
let exchange
|
|
let market
|
|
|
|
let learningSystem
|
|
let learningEngine
|
|
let sessionParameters
|
|
let dynamicIndicators
|
|
|
|
let machineLearningLibraryModuleObject
|
|
|
|
return thisObject
|
|
|
|
function initialize() {
|
|
learningSystem = TS.projects.foundations.globals.processVariables.VARIABLES_BY_PROCESS_INDEX_MAP.get(processIndex).SIMULATION_STATE.learningSystem
|
|
learningEngine = TS.projects.foundations.globals.processVariables.VARIABLES_BY_PROCESS_INDEX_MAP.get(processIndex).SIMULATION_STATE.learningEngine
|
|
sessionParameters = TS.projects.foundations.globals.processConstants.CONSTANTS_BY_PROCESS_INDEX_MAP.get(processIndex).SESSION_NODE.learningParameters
|
|
|
|
learningSystem.conditions = new Map()
|
|
learningSystem.formulas = new Map()
|
|
|
|
/* Adding Functions used elsewhere to Learning System Definition */
|
|
learningSystem.checkConditions = function (situation, passed) {
|
|
/* Given a Situation, we check if all children conditions are true or not */
|
|
for (let m = 0; m < situation.conditions.length; m++) {
|
|
|
|
let condition = situation.conditions[m]
|
|
let value = false
|
|
if (learningSystem.conditions.get(condition.id) !== undefined) {
|
|
value = learningSystem.conditions.get(condition.id)
|
|
}
|
|
if (value === true) {
|
|
learningSystem.highlights.push(condition.id)
|
|
}
|
|
else {
|
|
passed = false
|
|
}
|
|
learningSystem.values.push([condition.id, value])
|
|
}
|
|
return passed
|
|
}
|
|
|
|
learningSystem.evalConditions = function (startingNode, descendentOfNodeType) {
|
|
evalNode(startingNode, 'Conditions', descendentOfNodeType)
|
|
}
|
|
|
|
learningSystem.evalFormulas = function (startingNode, descendentOfNodeType) {
|
|
evalNode(startingNode, 'Formulas', descendentOfNodeType)
|
|
}
|
|
|
|
learningSystem.addError = function (errorDataArray) {
|
|
/*
|
|
This function adds to the array of error info a rate.
|
|
This rate will later help plotting the error at the
|
|
charts.
|
|
*/
|
|
let rate = learningEngine.learningCurrent.learningEpisode.candle.close.value
|
|
errorDataArray.push(rate)
|
|
learningSystem.errors.push(errorDataArray)
|
|
}
|
|
|
|
learningSystem.addWarning = function (warningDataArray) {
|
|
/*
|
|
This function adds to the array of warning info a rate.
|
|
This rate will later help plotting the warning at the
|
|
charts.
|
|
*/
|
|
let rate = learningEngine.learningCurrent.learningEpisode.candle.close.value
|
|
warningDataArray.push(rate)
|
|
learningSystem.warnings.push(warningDataArray)
|
|
}
|
|
|
|
learningSystem.addInfo = function (infoDataArray) {
|
|
/*
|
|
This function adds to the array of warning info a rate.
|
|
This rate will later help plotting the warning at the
|
|
charts.
|
|
*/
|
|
let rate = learningEngine.learningCurrent.learningEpisode.candle.close.value
|
|
infoDataArray.push(rate)
|
|
learningSystem.infos.push(infoDataArray)
|
|
}
|
|
|
|
/*
|
|
Initialize the Machine Learning Library
|
|
*/
|
|
|
|
let machineLearningLibrary = learningSystem.machineLearningLibrary
|
|
|
|
if (machineLearningLibrary === undefined) {
|
|
// TODO
|
|
}
|
|
/*
|
|
We will scan the project schema until we find the machine learning library
|
|
defined by the user at the UI.
|
|
*/
|
|
for (let i = 0; i < PROJECTS_SCHEMA.length; i++) {
|
|
let projectSchemaProject = PROJECTS_SCHEMA[i]
|
|
if (projectSchemaProject.name !== machineLearningLibrary.project) { continue }
|
|
|
|
for (let j = 0; j < projectSchemaProject.TS.botModules.length; j++) {
|
|
let botModuleDefinition = projectSchemaProject.TS.botModules[j]
|
|
|
|
/*
|
|
We match the type of the machine learning library with the name at the project schema
|
|
*/
|
|
if (botModuleDefinition.name === machineLearningLibrary.type) {
|
|
|
|
let project = TS.projects[machineLearningLibrary.project.toLowerCase()]
|
|
let botModule = project.botModules[botModuleDefinition.propertyName]
|
|
let moduleFunction = botModule[botModuleDefinition.functionName]
|
|
machineLearningLibraryModuleObject = moduleFunction(processIndex)
|
|
}
|
|
}
|
|
}
|
|
machineLearningLibraryModuleObject.initialize(onInitializeReady)
|
|
|
|
function onInitializeReady(err) {
|
|
if (err.result === TS.projects.foundations.globals.standardResponses.DEFAULT_OK_RESPONSE.result) {
|
|
|
|
} else {
|
|
|
|
}
|
|
}
|
|
|
|
function isItInside(elementWithTimestamp, elementWithBeginEnd) {
|
|
/*
|
|
The function is to allow in Conditions and Formulas to easily
|
|
know when an element with timestamp (like the ones of
|
|
single files) are inside the time range
|
|
of an element with a time range, like a candle.
|
|
*/
|
|
if (elementWithTimestamp.timestamp >= elementWithBeginEnd.begin && elementWithTimestamp.timestamp <= elementWithBeginEnd.end) {
|
|
return true
|
|
} else {
|
|
return false
|
|
}
|
|
}
|
|
}
|
|
|
|
function finalize() {
|
|
|
|
machineLearningLibraryModuleObject.finalize()
|
|
machineLearningLibraryModuleObject = undefined
|
|
|
|
chart = undefined
|
|
exchange = undefined
|
|
market = undefined
|
|
|
|
learningSystem.conditions = undefined
|
|
learningSystem.formulas = undefined
|
|
|
|
learningSystem.highlights = undefined
|
|
learningSystem.errors = undefined
|
|
learningSystem.warnings = undefined
|
|
learningSystem.infos = undefined
|
|
learningSystem.values = undefined
|
|
learningSystem.status = undefined
|
|
learningSystem.progress = undefined
|
|
learningSystem.running = undefined
|
|
|
|
learningSystem = undefined
|
|
learningEngine = undefined
|
|
sessionParameters = undefined
|
|
}
|
|
|
|
async function loadModel() {
|
|
await machineLearningLibraryModuleObject.loadModel()
|
|
}
|
|
|
|
async function saveModel() {
|
|
await machineLearningLibraryModuleObject.saveModel()
|
|
}
|
|
|
|
function mantain() {
|
|
machineLearningLibraryModuleObject.mantain()
|
|
}
|
|
|
|
function reset() {
|
|
machineLearningLibraryModuleObject.reset()
|
|
|
|
learningSystem.highlights = []
|
|
learningSystem.errors = []
|
|
learningSystem.warnings = []
|
|
learningSystem.infos = []
|
|
learningSystem.values = []
|
|
learningSystem.status = []
|
|
learningSystem.progress = []
|
|
learningSystem.running = []
|
|
learningSystem.announcements = []
|
|
}
|
|
|
|
function updateChart(pChart, pExchange, pMarket) {
|
|
/*
|
|
We need these 3 data structures to be a local objects
|
|
accessible while evaluating conditions and formulas.
|
|
*/
|
|
chart = pChart
|
|
exchange = pExchange
|
|
market = pMarket
|
|
|
|
machineLearningLibraryModuleObject.updateChart(pChart, pExchange, pMarket)
|
|
}
|
|
|
|
function buildDynamicIndicators() {
|
|
if (learningSystem.dynamicIndicators !== undefined) {
|
|
dynamicIndicators = {}
|
|
/* Eval Dynamic Indicators */
|
|
learningSystem.evalFormulas(learningSystem.dynamicIndicators, 'Indicator Function')
|
|
|
|
for (let i = 0; i < learningSystem.dynamicIndicators.indicatorFunctions.length; i++) {
|
|
let indicatorFunction = learningSystem.dynamicIndicators.indicatorFunctions[i]
|
|
if (indicatorFunction.formula === undefined) { return }
|
|
if (indicatorFunction.config.codeName === undefined) { return }
|
|
dynamicIndicators[indicatorFunction.config.codeName] = learningSystem.formulas.get(indicatorFunction.formula.id)
|
|
}
|
|
}
|
|
}
|
|
|
|
async function run() {
|
|
try {
|
|
/* Dynamic Indicators */
|
|
// buildDynamicIndicators() /* ***NOTE*** Would need to be implemented. */
|
|
|
|
/*
|
|
Initialize the Learning Algorithm
|
|
*/
|
|
await machineLearningLibraryModuleObject.run()
|
|
|
|
} catch (err) {
|
|
/*
|
|
If an error ocurred during execution, it was alreeady logged and
|
|
included at the errors array. That means we need to do nothing here,
|
|
just prevent the execution to be halted for not handling exceptions.
|
|
*/
|
|
if (typeof err === 'string' || err instanceof String) {
|
|
TS.projects.foundations.globals.loggerVariables.VARIABLES_BY_PROCESS_INDEX_MAP.get(processIndex).BOT_MAIN_LOOP_LOGGER_MODULE_OBJECT.write(MODULE_NAME, '[ERROR] runExecution -> err = ' + err)
|
|
}
|
|
if (err.stack !== undefined) {
|
|
TS.projects.foundations.globals.loggerVariables.VARIABLES_BY_PROCESS_INDEX_MAP.get(processIndex).BOT_MAIN_LOOP_LOGGER_MODULE_OBJECT.write(MODULE_NAME, '[ERROR] runExecution -> err = ' + err.stack)
|
|
}
|
|
}
|
|
}
|
|
|
|
function evalNode(node, evaluating, descendentOfNodeType, isDescendent) {
|
|
if (node === undefined) { return }
|
|
|
|
/* Verify if this node is decendent of the specified node type */
|
|
if (isDescendent !== true) {
|
|
if (node.type === descendentOfNodeType) {
|
|
isDescendent = true
|
|
}
|
|
}
|
|
|
|
/* Here we check if there is a codition to be evaluated */
|
|
if ((node.type === 'Condition' || node.type === 'Collection Condition') && evaluating === 'Conditions') {
|
|
/* We will eval this condition */
|
|
if (isDescendent === true) {
|
|
evalCondition(node)
|
|
}
|
|
}
|
|
|
|
/* Here we check if there is a formula to be evaluated */
|
|
if ((node.type === 'Feature Formula' || node.type === 'Label Formula') && evaluating === 'Formulas') {
|
|
if (node.code !== undefined) {
|
|
/* We will eval this formula */
|
|
if (isDescendent === true) {
|
|
evalFormula(node)
|
|
}
|
|
}
|
|
}
|
|
|
|
/* Now we go down through all this node children */
|
|
let schemaDocument = TS.projects.foundations.globals.taskConstants.APP_SCHEMA_MAP.get(node.project + '-' + node.type)
|
|
if (schemaDocument === undefined) { return }
|
|
|
|
if (schemaDocument.childrenNodesProperties !== undefined) {
|
|
let previousPropertyName // Since there are cases where there are many properties with the same name,because they can hold nodes of different types but only one at the time, we have to avoind counting each property of those as individual children.
|
|
for (let i = 0; i < schemaDocument.childrenNodesProperties.length; i++) {
|
|
let property = schemaDocument.childrenNodesProperties[i]
|
|
|
|
switch (property.type) {
|
|
case 'node': {
|
|
if (property.name !== previousPropertyName) {
|
|
if (node[property.name] !== undefined) {
|
|
evalNode(node[property.name], evaluating, descendentOfNodeType, isDescendent)
|
|
}
|
|
previousPropertyName = property.name
|
|
}
|
|
break
|
|
}
|
|
case 'array': {
|
|
if (node[property.name] !== undefined) {
|
|
let nodePropertyArray = node[property.name]
|
|
for (let m = 0; m < nodePropertyArray.length; m++) {
|
|
evalNode(nodePropertyArray[m], evaluating, descendentOfNodeType, isDescendent)
|
|
}
|
|
}
|
|
break
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
function evalCondition(node) {
|
|
let value
|
|
let errorMessage
|
|
let docs
|
|
/*
|
|
The code can be at the condition node if it was done with the Conditions Editor, or it can also be
|
|
at a Javascript Code node. If there is a Javascript object we will give it preference and take the code
|
|
from there. Otherwise we will take the code from the Condition node.
|
|
*/
|
|
let code = node.code
|
|
let nodeName = node.name
|
|
if (node.javascriptCode !== undefined) {
|
|
if (node.javascriptCode.code !== undefined) {
|
|
code = node.javascriptCode.code
|
|
nodeName = node.javascriptCode.name
|
|
}
|
|
}
|
|
|
|
try {
|
|
TS.projects.foundations.globals.loggerVariables.VARIABLES_BY_PROCESS_INDEX_MAP.get(processIndex).BOT_MAIN_LOOP_LOGGER_MODULE_OBJECT.write(MODULE_NAME, '[INFO] evalCondition -> ' + node.name + ' -> code = ' + code)
|
|
value = eval(code)
|
|
} catch (err) {
|
|
/*
|
|
One possible error is that the conditions references a .previous that is undefined. This
|
|
will not be considered an error.
|
|
*/
|
|
value = false
|
|
errorMessage = err.message
|
|
docs = {
|
|
project: 'Foundations',
|
|
category: 'Topic',
|
|
type: 'TS LF Learning Bot Error - Evaluating Condition Error',
|
|
placeholder: {}
|
|
}
|
|
TS.projects.education.utilities.docsFunctions.buildPlaceholder(docs, err, nodeName, code, undefined)
|
|
}
|
|
|
|
learningSystem.conditions.set(node.id, value)
|
|
|
|
if (value === true) {
|
|
learningSystem.highlights.push(node.id)
|
|
}
|
|
if (errorMessage !== undefined) {
|
|
learningSystem.addError([node.id, errorMessage, docs])
|
|
}
|
|
if (value !== undefined) {
|
|
learningSystem.values.push([node.id, value])
|
|
}
|
|
|
|
TS.projects.foundations.globals.loggerVariables.VARIABLES_BY_PROCESS_INDEX_MAP.get(processIndex).BOT_MAIN_LOOP_LOGGER_MODULE_OBJECT.write(MODULE_NAME, '[INFO] evalCondition -> value = ' + value)
|
|
TS.projects.foundations.globals.loggerVariables.VARIABLES_BY_PROCESS_INDEX_MAP.get(processIndex).BOT_MAIN_LOOP_LOGGER_MODULE_OBJECT.write(MODULE_NAME, '[INFO] evalCondition -> errorMessage = ' + errorMessage)
|
|
}
|
|
|
|
function evalFormula(node) {
|
|
let value
|
|
let errorMessage
|
|
let docs
|
|
|
|
try {
|
|
TS.projects.foundations.globals.loggerVariables.VARIABLES_BY_PROCESS_INDEX_MAP.get(processIndex).BOT_MAIN_LOOP_LOGGER_MODULE_OBJECT.write(MODULE_NAME, '[INFO] evalFormula -> ' + node.name + ' -> code = ' + node.code)
|
|
value = eval(node.code)
|
|
} catch (err) {
|
|
/*
|
|
One possible error is that the formula references a .previous that is undefined. This
|
|
will not be considered an error.
|
|
*/
|
|
value = 0
|
|
errorMessage = err.message
|
|
docs = {
|
|
project: 'Foundations',
|
|
category: 'Topic',
|
|
type: 'TS LF Learning Bot Error - Evaluating Formula Error',
|
|
placeholder: {}
|
|
}
|
|
TS.projects.education.utilities.docsFunctions.buildPlaceholder(docs, err, node.name, node.code, undefined)
|
|
}
|
|
|
|
if (errorMessage !== undefined) {
|
|
learningSystem.addError([node.id, errorMessage, docs])
|
|
TS.projects.foundations.globals.loggerVariables.VARIABLES_BY_PROCESS_INDEX_MAP.get(processIndex).BOT_MAIN_LOOP_LOGGER_MODULE_OBJECT.write(MODULE_NAME, '[INFO] evalFormula -> errorMessage = ' + errorMessage)
|
|
return
|
|
} else {
|
|
if (value !== undefined) {
|
|
if ((node.type === 'Feature Formula' || node.type === 'Label Formula') && isNaN(value)) {
|
|
|
|
docs = {
|
|
project: 'Foundations',
|
|
category: 'Topic',
|
|
type: 'TS LF Learning Bot Error - Formula Value Not A Number',
|
|
placeholder: {}
|
|
}
|
|
TS.projects.education.utilities.docsFunctions.buildPlaceholder(docs, undefined, node.name, node.code, undefined, value)
|
|
|
|
learningSystem.addError([node.id, 'Formula needs to return a numeric value.', docs])
|
|
return
|
|
}
|
|
|
|
learningSystem.values.push([node.id, value])
|
|
learningSystem.formulas.set(node.id, value)
|
|
}
|
|
}
|
|
|
|
TS.projects.foundations.globals.loggerVariables.VARIABLES_BY_PROCESS_INDEX_MAP.get(processIndex).BOT_MAIN_LOOP_LOGGER_MODULE_OBJECT.write(MODULE_NAME, '[INFO] evalFormula -> value = ' + value)
|
|
}
|
|
} |