aiecs 1.5.1__py3-none-any.whl

aiecs/tools/statistics/ai_insight_generator_tool.py

@@ -0,0 +1,505 @@
+"""
+AI Insight Generator Tool - AI-driven insight discovery and pattern detection
+
+This tool provides advanced insight generation with:
+- Pattern discovery and anomaly detection
+- Trend analysis and forecasting
+- Actionable insight generation
+- Integration with research_tool reasoning methods
+- AI-powered analysis (placeholder for future enhancement)
+"""
+
+import logging
+from typing import Dict, Any, List, Optional, Union
+from enum import Enum
+from datetime import datetime
+
+import pandas as pd
+import numpy as np
+from scipy import stats as scipy_stats
+from pydantic import BaseModel, Field, ConfigDict
+
+from aiecs.tools.base_tool import BaseTool
+from aiecs.tools import register_tool
+
+
+class InsightType(str, Enum):
+    """Types of insights to generate"""
+
+    PATTERN = "pattern"
+    ANOMALY = "anomaly"
+    TREND = "trend"
+    CORRELATION = "correlation"
+    SEGMENTATION = "segmentation"
+    CAUSATION = "causation"
+
+
+class InsightGeneratorError(Exception):
+    """Base exception for Insight Generator errors"""
+
+
+class InsightGenerationError(InsightGeneratorError):
+    """Raised when insight generation fails"""
+
+
+@register_tool("ai_insight_generator")
+class AIInsightGeneratorTool(BaseTool):
+    """
+    AI-powered insight generation tool that can:
+    1. Discover hidden patterns in data
+    2. Generate actionable insights
+    3. Detect anomalies and outliers
+    4. Predict trends and forecast
+    5. Apply reasoning methods (Mill's methods, induction, deduction)
+
+    Integrates with research_tool for reasoning capabilities.
+    """
+
+    # Configuration schema
+    class Config(BaseModel):
+        """Configuration for the AI insight generator tool"""
+
+        model_config = ConfigDict(env_prefix="AI_INSIGHT_GENERATOR_")
+
+        min_confidence: float = Field(
+            default=0.7,
+            description="Minimum confidence threshold for insights",
+        )
+        anomaly_std_threshold: float = Field(
+            default=3.0,
+            description="Standard deviation threshold for anomaly detection",
+        )
+        correlation_threshold: float = Field(
+            default=0.5,
+            description="Correlation threshold for significant relationships",
+        )
+        enable_reasoning: bool = Field(
+            default=True,
+            description="Whether to enable reasoning methods integration",
+        )
+
+    def __init__(self, config: Optional[Dict[str, Any]] = None):
+        """Initialize AI Insight Generator Tool"""
+        super().__init__(config)
+
+        # Parse configuration
+        self.config = self.Config(**(config or {}))
+
+        self.logger = logging.getLogger(__name__)
+        if not self.logger.handlers:
+            handler = logging.StreamHandler()
+            handler.setFormatter(logging.Formatter("%(asctime)s %(levelname)s %(message)s"))
+            self.logger.addHandler(handler)
+        self.logger.setLevel(logging.INFO)
+
+        self._init_external_tools()
+
+    def _init_external_tools(self):
+        """Initialize external task tools"""
+        self.external_tools = {}
+
+        # Initialize ResearchTool for reasoning methods
+        try:
+            from aiecs.tools.task_tools.research_tool import ResearchTool
+
+            self.external_tools["research"] = ResearchTool()
+            self.logger.info("ResearchTool initialized successfully")
+        except ImportError:
+            self.logger.warning("ResearchTool not available")
+            self.external_tools["research"] = None
+
+        # Initialize StatisticalAnalyzerTool
+        try:
+            from aiecs.tools.statistics.statistical_analyzer_tool import (
+                StatisticalAnalyzerTool,
+            )
+
+            self.external_tools["stats_analyzer"] = StatisticalAnalyzerTool()
+            self.logger.info("StatisticalAnalyzerTool initialized successfully")
+        except ImportError:
+            self.logger.warning("StatisticalAnalyzerTool not available")
+            self.external_tools["stats_analyzer"] = None
+
+    # Schema definitions
+    class GenerateInsightsSchema(BaseModel):
+        """Schema for generate_insights operation"""
+
+        data: Union[Dict[str, Any], List[Dict[str, Any]]] = Field(description="Data to analyze")
+        analysis_results: Optional[Dict[str, Any]] = Field(
+            default=None, description="Previous analysis results"
+        )
+        insight_types: Optional[List[InsightType]] = Field(
+            default=None, description="Types of insights to generate"
+        )
+        min_confidence: float = Field(default=0.7, description="Minimum confidence threshold")
+
+    class DiscoverPatternsSchema(BaseModel):
+        """Schema for discover_patterns operation"""
+
+        data: Union[Dict[str, Any], List[Dict[str, Any]]] = Field(
+            description="Data for pattern discovery"
+        )
+        pattern_types: Optional[List[str]] = Field(
+            default=None, description="Specific pattern types"
+        )
+
+    class DetectAnomaliesSchema(BaseModel):
+        """Schema for detect_anomalies operation"""
+
+        data: Union[Dict[str, Any], List[Dict[str, Any]]] = Field(
+            description="Data for anomaly detection"
+        )
+        columns: Optional[List[str]] = Field(default=None, description="Columns to check")
+        threshold: float = Field(default=3.0, description="Standard deviation threshold")
+
+    def generate_insights(
+        self,
+        data: Union[Dict[str, Any], List[Dict[str, Any]], pd.DataFrame],
+        analysis_results: Optional[Dict[str, Any]] = None,
+        insight_types: Optional[List[InsightType]] = None,
+        min_confidence: float = 0.7,
+    ) -> Dict[str, Any]:
+        """
+        Generate AI-powered insights from data and analysis results.
+
+        Args:
+            data: Data to analyze
+            analysis_results: Previous analysis results to incorporate
+            insight_types: Specific types of insights to generate (all if None)
+            min_confidence: Minimum confidence threshold for insights
+
+        Returns:
+            Dict containing:
+                - insights: List of generated insights
+                - summary: Overall summary
+                - priority_insights: Top priority insights
+        """
+        try:
+            df = self._to_dataframe(data)
+
+            self.logger.info(f"Generating insights from data with {len(df)} rows")
+
+            # Default to all insight types
+            if insight_types is None:
+                insight_types = list(InsightType)
+
+            insights = []
+
+            # Generate different types of insights
+            if InsightType.PATTERN in insight_types:
+                pattern_insights = self._discover_patterns_internal(df)
+                insights.extend(pattern_insights)
+
+            if InsightType.ANOMALY in insight_types:
+                anomaly_insights = self._detect_anomalies_internal(df)
+                insights.extend(anomaly_insights)
+
+            if InsightType.TREND in insight_types:
+                trend_insights = self._analyze_trends_internal(df)
+                insights.extend(trend_insights)
+
+            if InsightType.CORRELATION in insight_types:
+                correlation_insights = self._analyze_correlations_internal(df)
+                insights.extend(correlation_insights)
+
+            if InsightType.CAUSATION in insight_types and self.config.enable_reasoning:
+                causation_insights = self._analyze_causation_internal(df)
+                insights.extend(causation_insights)
+
+            # Filter by confidence
+            filtered_insights = [i for i in insights if i.get("confidence", 0) >= min_confidence]
+
+            # Prioritize insights
+            priority_insights = self._prioritize_insights(filtered_insights)
+
+            # Generate summary
+            summary = self._generate_insight_summary(filtered_insights)
+
+            return {
+                "insights": filtered_insights,
+                "summary": summary,
+                "priority_insights": priority_insights[:5],
+                "total_insights": len(filtered_insights),
+                "timestamp": datetime.now().isoformat(),
+            }
+
+        except Exception as e:
+            self.logger.error(f"Error generating insights: {e}")
+            raise InsightGenerationError(f"Insight generation failed: {e}")
+
+    def discover_patterns(
+        self,
+        data: Union[Dict[str, Any], List[Dict[str, Any]], pd.DataFrame],
+        pattern_types: Optional[List[str]] = None,
+    ) -> Dict[str, Any]:
+        """
+        Discover patterns in data.
+
+        Args:
+            data: Data for pattern discovery
+            pattern_types: Specific pattern types to look for
+
+        Returns:
+            Dict containing discovered patterns
+        """
+        try:
+            df = self._to_dataframe(data)
+            patterns = self._discover_patterns_internal(df)
+
+            return {"patterns": patterns, "total_patterns": len(patterns)}
+
+        except Exception as e:
+            self.logger.error(f"Error discovering patterns: {e}")
+            raise InsightGenerationError(f"Pattern discovery failed: {e}")
+
+    def detect_anomalies(
+        self,
+        data: Union[Dict[str, Any], List[Dict[str, Any]], pd.DataFrame],
+        columns: Optional[List[str]] = None,
+        threshold: float = 3.0,
+    ) -> Dict[str, Any]:
+        """
+        Detect anomalies in data.
+
+        Args:
+            data: Data for anomaly detection
+            columns: Columns to check (all numeric if None)
+            threshold: Standard deviation threshold
+
+        Returns:
+            Dict containing detected anomalies
+        """
+        try:
+            df = self._to_dataframe(data)
+            anomalies = self._detect_anomalies_internal(df, columns, threshold)
+
+            return {"anomalies": anomalies, "total_anomalies": len(anomalies)}
+
+        except Exception as e:
+            self.logger.error(f"Error detecting anomalies: {e}")
+            raise InsightGenerationError(f"Anomaly detection failed: {e}")
+
+    # Internal insight generation methods
+
+    def _to_dataframe(self, data: Union[Dict, List, pd.DataFrame]) -> pd.DataFrame:
+        """Convert data to DataFrame"""
+        if isinstance(data, pd.DataFrame):
+            return data
+        elif isinstance(data, list):
+            return pd.DataFrame(data)
+        elif isinstance(data, dict):
+            return pd.DataFrame([data])
+        else:
+            raise InsightGenerationError(f"Unsupported data type: {type(data)}")
+
+    def _discover_patterns_internal(self, df: pd.DataFrame) -> List[Dict[str, Any]]:
+        """Discover patterns in data"""
+        patterns = []
+
+        # Distribution patterns
+        numeric_cols = df.select_dtypes(include=[np.number]).columns
+        for col in numeric_cols:
+            series = df[col].dropna()
+            if len(series) > 0:
+                skewness = series.skew()
+                if abs(skewness) > 1:
+                    patterns.append(
+                        {
+                            "type": InsightType.PATTERN.value,
+                            "title": f"Skewed Distribution in {col}",
+                            "description": f"Column {col} shows {'positive' if skewness > 0 else 'negative'} skew ({skewness:.2f})",
+                            "confidence": 0.85,
+                            "impact": "medium",
+                            "evidence": {
+                                "skewness": float(skewness),
+                                "column": col,
+                            },
+                        }
+                    )
+
+        # Categorical patterns
+        categorical_cols = df.select_dtypes(include=["object", "category"]).columns
+        for col in categorical_cols:
+            value_counts = df[col].value_counts()
+            if len(value_counts) > 0:
+                top_percentage = value_counts.iloc[0] / len(df) * 100
+                if top_percentage > 50:
+                    patterns.append(
+                        {
+                            "type": InsightType.PATTERN.value,
+                            "title": f"Dominant Category in {col}",
+                            "description": f"'{value_counts.index[0]}' accounts for {top_percentage:.1f}% of {col}",
+                            "confidence": 0.9,
+                            "impact": "high",
+                            "evidence": {
+                                "dominant_value": str(value_counts.index[0]),
+                                "percentage": float(top_percentage),
+                            },
+                        }
+                    )
+
+        return patterns
+
+    def _detect_anomalies_internal(
+        self,
+        df: pd.DataFrame,
+        columns: Optional[List[str]] = None,
+        threshold: float = 3.0,
+    ) -> List[Dict[str, Any]]:
+        """Detect anomalies using statistical methods"""
+        anomalies = []
+
+        numeric_cols = (
+            columns if columns else df.select_dtypes(include=[np.number]).columns.tolist()
+        )
+
+        for col in numeric_cols:
+            if col not in df.columns:
+                continue
+
+            series = df[col].dropna()
+            if len(series) == 0 or series.std() == 0:
+                continue
+
+            # Z-score method
+            z_scores = np.abs((series - series.mean()) / series.std())
+            anomaly_count = (z_scores > threshold).sum()
+
+            if anomaly_count > 0:
+                anomaly_percentage = anomaly_count / len(series) * 100
+                anomalies.append(
+                    {
+                        "type": InsightType.ANOMALY.value,
+                        "title": f"Anomalies Detected in {col}",
+                        "description": f"Found {anomaly_count} anomalous values ({anomaly_percentage:.2f}%) in {col}",
+                        "confidence": 0.8,
+                        "impact": ("high" if anomaly_percentage > 5 else "medium"),
+                        "evidence": {
+                            "column": col,
+                            "anomaly_count": int(anomaly_count),
+                            "percentage": float(anomaly_percentage),
+                            "threshold": threshold,
+                        },
+                        "recommendation": "Investigate and consider handling these outliers",
+                    }
+                )
+
+        return anomalies
+
+    def _analyze_trends_internal(self, df: pd.DataFrame) -> List[Dict[str, Any]]:
+        """Analyze trends in data"""
+        trends = []
+
+        numeric_cols = df.select_dtypes(include=[np.number]).columns
+
+        for col in numeric_cols:
+            series = df[col].dropna()
+            if len(series) < 3:
+                continue
+
+            # Calculate trend using linear regression
+            x = np.arange(len(series))
+            y = series.values
+
+            if len(x) > 0 and len(y) > 0:
+                slope, intercept, r_value, p_value, std_err = scipy_stats.linregress(x, y)
+
+                if abs(r_value) > 0.5 and p_value < 0.05:
+                    trend_direction = "increasing" if slope > 0 else "decreasing"
+                    trends.append(
+                        {
+                            "type": InsightType.TREND.value,
+                            "title": f"{trend_direction.capitalize()} Trend in {col}",
+                            "description": f"Column {col} shows a {trend_direction} trend (R²={r_value**2:.3f})",
+                            "confidence": float(abs(r_value)),
+                            "impact": ("high" if abs(r_value) > 0.7 else "medium"),
+                            "evidence": {
+                                "column": col,
+                                "slope": float(slope),
+                                "r_squared": float(r_value**2),
+                                "p_value": float(p_value),
+                            },
+                        }
+                    )
+
+        return trends
+
+    def _analyze_correlations_internal(self, df: pd.DataFrame) -> List[Dict[str, Any]]:
+        """Analyze correlations between variables"""
+        correlations = []
+
+        numeric_df = df.select_dtypes(include=[np.number])
+        if numeric_df.shape[1] < 2:
+            return correlations
+
+        corr_matrix = numeric_df.corr()
+
+        for i in range(len(corr_matrix.columns)):
+            for j in range(i + 1, len(corr_matrix.columns)):
+                corr_value = corr_matrix.iloc[i, j]
+
+                if abs(corr_value) > self.config.correlation_threshold:
+                    col1 = corr_matrix.columns[i]
+                    col2 = corr_matrix.columns[j]
+
+                    strength = "strong" if abs(corr_value) > 0.7 else "moderate"
+                    direction = "positive" if corr_value > 0 else "negative"
+
+                    correlations.append(
+                        {
+                            "type": InsightType.CORRELATION.value,
+                            "title": f"{strength.capitalize()} {direction} correlation",
+                            "description": f"{col1} and {col2} show {strength} {direction} correlation ({corr_value:.3f})",
+                            "confidence": float(abs(corr_value)),
+                            "impact": ("high" if abs(corr_value) > 0.7 else "medium"),
+                            "evidence": {
+                                "variable1": col1,
+                                "variable2": col2,
+                                "correlation": float(corr_value),
+                            },
+                            "recommendation": "Consider investigating causal relationship",
+                        }
+                    )
+
+        return correlations
+
+    def _analyze_causation_internal(self, df: pd.DataFrame) -> List[Dict[str, Any]]:
+        """Analyze potential causal relationships using reasoning methods"""
+        causations = []
+
+        # Use research tool for Mill's methods if available
+        if self.external_tools.get("research"):
+            # Placeholder for causal analysis using Mill's methods
+            # This would require domain knowledge and proper case structures
+            self.logger.info(
+                "Causal analysis with reasoning methods available but requires domain-specific setup"
+            )
+
+        return causations
+
+    def _prioritize_insights(self, insights: List[Dict[str, Any]]) -> List[Dict[str, Any]]:
+        """Prioritize insights by confidence and impact"""
+        impact_scores = {"high": 3, "medium": 2, "low": 1}
+
+        def priority_score(insight):
+            confidence = insight.get("confidence", 0.5)
+            impact = impact_scores.get(insight.get("impact", "low"), 1)
+            return confidence * impact
+
+        return sorted(insights, key=priority_score, reverse=True)
+
+    def _generate_insight_summary(self, insights: List[Dict[str, Any]]) -> str:
+        """Generate summary of insights"""
+        if not insights:
+            return "No significant insights found in the data."
+
+        type_counts = {}
+        for insight in insights:
+            insight_type = insight.get("type", "unknown")
+            type_counts[insight_type] = type_counts.get(insight_type, 0) + 1
+
+        summary_parts = [f"Generated {len(insights)} insights:"]
+        for itype, count in type_counts.items():
+            summary_parts.append(f"{count} {itype} insights")
+
+        return "; ".join(summary_parts)