Macro Engine with AI Predictions

This case study shows how Ascoos OS can execute macros based on AI predictions, combining logistic regression with a DSL (Domain-Specific Language). It uses the classes TArtificialIntelligenceHandler, AbstractDslAstBuilder and AstMacroTranslator to train a model, parse rules into an AST, translate them into macros and execute them dynamically.

Purpose

• Train a model using `trainLogisticRegression()`
• Build an AST from DSL with `AbstractDslAstBuilder`
• Translate the AST into executable macros with `AstMacroTranslator`
• Execute macros when predictions meet a condition

Core Ascoos OS Classes

• TArtificialIntelligenceHandler Train and predict logistic regression models
• AbstractDslAstBuilder Parse DSL statements into an Abstract Syntax Tree
• AstMacroTranslator Convert AST nodes into a container of callbacks for each macro

File Structure

This case study is implemented in a single PHP script: - macro_decision_engine.php Contains data loading, model training, DSL parsing, translation and macro execution.

Prerequisites

1. PHP ? 8.2
2. Ascoos OS installed. If you?re using ASCOOS Web Extended Studio (AWES) 26, it?s pre-installed.

Getting Started

1. Adjust your training data (`$X`, `$y`) as needed.

2. Run the script via your web server:
   https://localhost/aos/examples/case-studies/ai/macro_decision_engine/macro_decision_engine.php
   

DSL Example

WHEN predict(user.features) > 0.5 THEN
    LOG "User is eligible"
    ENABLE MODULE "AdvancedAnalytics"

Execution Flow

1. `TArtificialIntelligenceHandler` trains a logistic regression model with `$X` and `$y`.
2. `AbstractDslAstBuilder` transforms the DSL script into AST nodes.
3. `AstMacroTranslator` maps each node to a callback: - `LOG` ? print a message - `ENABLE MODULE` ? enable a specific module - `predict` ? call `predictLogisticRegression()`
4. `TMacroHandler` executes commands only if the prediction exceeds the threshold (`> 0.5`).

Code Example

// Train the model
$ai    = new TArtificialIntelligenceHandler();
$model = $ai->trainLogisticRegression($X, $y);

// Define the DSL
$dsl = <<<DSL
WHEN predict(user.features) > 0.5 THEN
    LOG "User is eligible"
    ENABLE MODULE "AdvancedAnalytics"
DSL;

// Build AST & translate
$astBuilder     = new class extends AbstractDslAstBuilder {};
$ast            = $astBuilder->buildAst($dsl);
$translator     = new class([...]) extends AstMacroTranslator {};
$macroContainer = $translator->translateAst($ast);

// Execute based on user features
$user = ['features' => [1, 1, 0]];
$macroContainer->executeIfTrue($user);

Expected Output

If the prediction predict([1,1,0]) > 0.5:

User is eligible
Module enabled: AdvancedAnalytics

Resources

• Ascoos OS Documentation
• GitHub Repository
• AWES
• BootLib
• phpBCL8

Contribution

Want to contribute to this case study? Fork the repo, add new macros or DSL enhancements in macro_decision_engine.php and submit a pull request. See CONTRIBUTING.md for guidelines.

License

This case study is covered by the Ascoos General License (AGL). See LICENSE.