115 lines
3.5 KiB
Dart
115 lines
3.5 KiB
Dart
import 'dart:math';
|
|
import '../models/enums.dart';
|
|
import '../models/report.dart';
|
|
|
|
/// Service for generating deterministic AI suggestions for reports
|
|
class MockAIService {
|
|
/// Generate a deterministic seed based on report parameters
|
|
static int _generateSeed(String id, String createdAt, double lat, double lng, int? photoSizeBytes) {
|
|
final combined = '$id$createdAt$lat$lng${photoSizeBytes ?? 0}';
|
|
var hash = 0;
|
|
for (var i = 0; i < combined.length; i++) {
|
|
hash = ((hash << 5) - hash) + combined.codeUnitAt(i);
|
|
hash = hash & hash; // Convert to 32-bit integer
|
|
}
|
|
return hash.abs();
|
|
}
|
|
|
|
/// Generate AI suggestion for a report
|
|
static AISuggestion generateSuggestion({
|
|
required String id,
|
|
required String createdAt,
|
|
required double lat,
|
|
required double lng,
|
|
int? photoSizeBytes,
|
|
}) {
|
|
final seed = _generateSeed(id, createdAt, lat, lng, photoSizeBytes);
|
|
final random = Random(seed);
|
|
|
|
// Category selection with weighted probabilities
|
|
final categoryWeights = {
|
|
Category.pothole: 0.35,
|
|
Category.trash: 0.25,
|
|
Category.streetlight: 0.15,
|
|
Category.signage: 0.10,
|
|
Category.drainage: 0.10,
|
|
Category.other: 0.05,
|
|
};
|
|
|
|
// Apply heuristics based on image dimensions (if available)
|
|
final aspectRatio = photoSizeBytes != null ? (random.nextDouble() * 2) : 1.0;
|
|
if (aspectRatio > 1.2) {
|
|
// Wide image - likely signage
|
|
categoryWeights[Category.signage] = categoryWeights[Category.signage]! * 2;
|
|
categoryWeights[Category.pothole] = categoryWeights[Category.pothole]! * 0.5;
|
|
}
|
|
|
|
// Select category based on weights
|
|
final categoryRand = random.nextDouble();
|
|
double cumulative = 0.0;
|
|
Category selectedCategory = Category.pothole;
|
|
|
|
for (final entry in categoryWeights.entries) {
|
|
cumulative += entry.value;
|
|
if (categoryRand <= cumulative) {
|
|
selectedCategory = entry.key;
|
|
break;
|
|
}
|
|
}
|
|
|
|
// Severity selection with weighted probabilities
|
|
final severityWeights = {
|
|
Severity.medium: 0.45,
|
|
Severity.high: 0.30,
|
|
Severity.low: 0.25,
|
|
};
|
|
|
|
// Apply location accuracy heuristic
|
|
final accuracy = random.nextDouble() * 50; // Simulate accuracy 0-50m
|
|
final isNight = random.nextBool(); // Simulate night time
|
|
|
|
if (accuracy <= 10 && isNight) {
|
|
// High accuracy at night - bump high severity
|
|
severityWeights[Severity.high] = severityWeights[Severity.high]! * 1.5;
|
|
severityWeights[Severity.medium] = severityWeights[Severity.medium]! * 0.8;
|
|
}
|
|
|
|
// Select severity based on weights
|
|
final severityRand = random.nextDouble();
|
|
cumulative = 0.0;
|
|
Severity selectedSeverity = Severity.medium;
|
|
|
|
for (final entry in severityWeights.entries) {
|
|
cumulative += entry.value;
|
|
if (severityRand <= cumulative) {
|
|
selectedSeverity = entry.key;
|
|
break;
|
|
}
|
|
}
|
|
|
|
// Generate confidence score (0.6 - 0.9)
|
|
final confidence = 0.6 + (random.nextDouble() * 0.3);
|
|
|
|
return AISuggestion(
|
|
category: selectedCategory,
|
|
severity: selectedSeverity,
|
|
confidence: confidence,
|
|
);
|
|
}
|
|
|
|
/// Check if the AI suggestion is reliable enough to use
|
|
static bool isSuggestionReliable(AISuggestion suggestion) {
|
|
return suggestion.confidence >= 0.7;
|
|
}
|
|
|
|
/// Get confidence level description
|
|
static String getConfidenceDescription(double confidence) {
|
|
if (confidence >= 0.8) {
|
|
return 'High confidence';
|
|
} else if (confidence >= 0.7) {
|
|
return 'Medium confidence';
|
|
} else {
|
|
return 'Low confidence';
|
|
}
|
|
}
|
|
} |