\r

Data Quality Check for Construction\r

\r

Overview\r

\r Based on DDC methodology (Chapter 2.6), this skill provides comprehensive data quality assessment for construction projects. Poor data quality leads to poor decisions - validate early, validate often.\r \r Book Reference: "Требования к качеству данных и его обеспечение" / "Data Quality Requirements"\r \r

"Качество данных определяется пятью ключевыми метриками: полнота, точность, согласованность, своевременность и достоверность."\r — DDC Book, Chapter 2.6\r \r

Quick Start\r

\r

import pandas as pd\r
\r
# Load construction data\r
df = pd.read_excel("bim_export.xlsx")\r
\r
# Quick quality check\r
quality_score = {\r
    'completeness': (1 - df.isnull().sum().sum() / df.size) * 100,\r
    'unique_ids': df['ElementId'].nunique() == len(df),\r
    'valid_volumes': (df['Volume_m3'] >= 0).all()\r
}\r
\r
print(f"Completeness: {quality_score['completeness']:.1f}%")\r
print(f"Unique IDs: {quality_score['unique_ids']}")\r
print(f"Valid volumes: {quality_score['valid_volumes']}")\r
```\r
\r
## Data Quality Dimensions\r
\r
### The 5 Quality Metrics\r
\r
```python\r
import pandas as pd\r
import numpy as np\r
import re\r
from datetime import datetime, timedelta\r
\r
class DataQualityChecker:\r
    """Comprehensive data quality assessment for construction data"""\r
\r
    def __init__(self, df):\r
        self.df = df.copy()\r
        self.results = {}\r
        self.issues = []\r
\r
    def check_completeness(self, required_columns=None):\r
        """Check for missing values (Полнота)"""\r
        if required_columns is None:\r
            required_columns = self.df.columns.tolist()\r
\r
        completeness = {}\r
        for col in required_columns:\r
            if col in self.df.columns:\r
                non_null = self.df[col].notna().sum()\r
                total = len(self.df)\r
                completeness[col] = (non_null / total) * 100\r
            else:\r
                completeness[col] = 0\r
                self.issues.append(f"Missing required column: {col}")\r
\r
        overall = np.mean(list(completeness.values()))\r
\r
        self.results['completeness'] = {\r
            'by_column': completeness,\r
            'overall': overall,\r
            'threshold': 95,\r
            'passed': overall >= 95\r
        }\r
\r
        return self.results['completeness']\r
\r
    def check_accuracy(self, rules=None):\r
        """Check data accuracy against rules (Точность)"""\r
        if rules is None:\r
            # Default construction data rules\r
            rules = {\r
                'Volume_m3': {'min': 0, 'max': 10000},\r
                'Area_m2': {'min': 0, 'max': 100000},\r
                'Weight_kg': {'min': 0, 'max': 1000000},\r
                'Cost': {'min': 0, 'max': 100000000}\r
            }\r
\r
        accuracy = {}\r
        for col, bounds in rules.items():\r
            if col in self.df.columns:\r
                valid = self.df[col].between(\r
                    bounds.get('min', -np.inf),\r
                    bounds.get('max', np.inf)\r
                ).sum()\r
                total = self.df[col].notna().sum()\r
                accuracy[col] = (valid / total * 100) if total > 0 else 100\r
\r
                # Log invalid values\r
                invalid_count = total - valid\r
                if invalid_count > 0:\r
                    self.issues.append(\r
                        f"{col}: {invalid_count} values outside range [{bounds.get('min')}, {bounds.get('max')}]"\r
                    )\r
\r
        overall = np.mean(list(accuracy.values())) if accuracy else 100\r
\r
        self.results['accuracy'] = {\r
            'by_column': accuracy,\r
            'overall': overall,\r
            'threshold': 98,\r
            'passed': overall >= 98\r
        }\r
\r
        return self.results['accuracy']\r
\r
    def check_consistency(self, unique_cols=None, relationship_rules=None):\r
        """Check data consistency (Согласованность)"""\r
        consistency = {}\r
\r
        # Check unique columns\r
        if unique_cols is None:\r
            unique_cols = ['ElementId']\r
\r
        for col in unique_cols:\r
            if col in self.df.columns:\r
                is_unique = self.df[col].nunique() == len(self.df)\r
                consistency[f'{col}_unique'] = 100 if is_unique else \\r
                    (self.df[col].nunique() / len(self.df) * 100)\r
\r
                if not is_unique:\r
                    duplicates = self.df[self.df[col].duplicated()][col].unique()\r
                    self.issues.append(f"Duplicate {col}: {len(duplicates)} duplicates found")\r
\r
        # Check cross-field relationships\r
        if relationship_rules is None:\r
            relationship_rules = [\r
                ('End_Date', '>=', 'Start_Date'),\r
                ('Gross_Volume', '>=', 'Net_Volume')\r
            ]\r
\r
        for col1, op, col2 in relationship_rules:\r
            if col1 in self.df.columns and col2 in self.df.columns:\r
                if op == '>=':\r
                    valid = (self.df[col1] >= self.df[col2]).sum()\r
                elif op == '>':\r
                    valid = (self.df[col1] > self.df[col2]).sum()\r
                elif op == '==':\r
                    valid = (self.df[col1] == self.df[col2]).sum()\r
\r
                total = self.df[[col1, col2]].notna().all(axis=1).sum()\r
                consistency[f'{col1}_{op}_{col2}'] = (valid / total * 100) if total > 0 else 100\r
\r
        overall = np.mean(list(consistency.values())) if consistency else 100\r
\r
        self.results['consistency'] = {\r
            'checks': consistency,\r
            'overall': overall,\r
            'threshold': 99,\r
            'passed': overall >= 99\r
        }\r
\r
        return self.results['consistency']\r
\r
    def check_timeliness(self, date_col='Modified_Date', max_age_days=30):\r
        """Check data timeliness (Своевременность)"""\r
        if date_col not in self.df.columns:\r
            self.results['timeliness'] = {\r
                'overall': None,\r
                'message': f'Column {date_col} not found'\r
            }\r
            return self.results['timeliness']\r
\r
        dates = pd.to_datetime(self.df[date_col], errors='coerce')\r
        cutoff = datetime.now() - timedelta(days=max_age_days)\r
\r
        recent = (dates >= cutoff).sum()\r
        total = dates.notna().sum()\r
        timeliness_pct = (recent / total * 100) if total > 0 else 0\r
\r
        oldest = dates.min()\r
        newest = dates.max()\r
        avg_age = (datetime.now() - dates.mean()).days if dates.notna().any() else None\r
\r
        self.results['timeliness'] = {\r
            'recent_percentage': timeliness_pct,\r
            'oldest_record': oldest,\r
            'newest_record': newest,\r
            'average_age_days': avg_age,\r
            'threshold': 80,\r
            'passed': timeliness_pct >= 80\r
        }\r
\r
        return self.results['timeliness']\r
\r
    def check_validity(self, patterns=None):\r
        """Check data validity with regex patterns (Достоверность)"""\r
        if patterns is None:\r
            patterns = {\r
                'ElementId': r'^[A-Z]{1,3}\d{3,6}$',  # e.g., W001, FL12345\r
                'Level': r'^Level\s*\d+$|^L\d+$|^Уровень\s*\d+$',\r
                'Email': r'^[\w\.-]+@[\w\.-]+\.\w+$',\r
                'Phone': r'^\+?\d{10,15}$'\r
            }\r
\r
        validity = {}\r
        for col, pattern in patterns.items():\r
            if col in self.df.columns:\r
                non_null = self.df[col].dropna()\r
                if len(non_null) > 0:\r
                    matches = non_null.astype(str).str.match(pattern).sum()\r
                    validity[col] = (matches / len(non_null) * 100)\r
\r
                    invalid = len(non_null) - matches\r
                    if invalid > 0:\r
                        self.issues.append(f"{col}: {invalid} values don't match pattern")\r
                else:\r
                    validity[col] = 100\r
\r
        overall = np.mean(list(validity.values())) if validity else 100\r
\r
        self.results['validity'] = {\r
            'by_column': validity,\r
            'overall': overall,\r
            'threshold': 95,\r
            'passed': overall >= 95\r
        }\r
\r
        return self.results['validity']\r
\r
    def run_full_check(self):\r
        """Run all quality checks"""\r
        self.check_completeness()\r
        self.check_accuracy()\r
        self.check_consistency()\r
        self.check_timeliness()\r
        self.check_validity()\r
\r
        # Calculate overall score\r
        scores = []\r
        for metric in ['completeness', 'accuracy', 'consistency', 'validity']:\r
            if metric in self.results and self.results[metric].get('overall'):\r
                scores.append(self.results[metric]['overall'])\r
\r
        self.results['overall_score'] = np.mean(scores) if scores else 0\r
        self.results['grade'] = self._calculate_grade(self.results['overall_score'])\r
        self.results['issues'] = self.issues\r
\r
        return self.results\r
\r
    def _calculate_grade(self, score):\r
        """Calculate quality grade"""\r
        if score >= 98:\r
            return 'A+'\r
        elif score >= 95:\r
            return 'A'\r
        elif score >= 90:\r
            return 'B'\r
        elif score >= 80:\r
            return 'C'\r
        elif score >= 70:\r
            return 'D'\r
        else:\r
            return 'F'\r
\r
    def generate_report(self):\r
        """Generate quality report"""\r
        if not self.results:\r
            self.run_full_check()\r
\r
        report = []\r
        report.append("=" * 60)\r
        report.append("DATA QUALITY REPORT")\r
        report.append("=" * 60)\r
        report.append(f"Records analyzed: {len(self.df)}")\r
        report.append(f"Columns: {len(self.df.columns)}")\r
        report.append("")\r
        report.append(f"OVERALL SCORE: {self.results['overall_score']:.1f}% (Grade: {self.results['grade']})")\r
        report.append("")\r
        report.append("-" * 60)\r
\r
        # Detail by dimension\r
        for metric in ['completeness', 'accuracy', 'consistency', 'validity', 'timeliness']:\r
            if metric in self.results:\r
                r = self.results[metric]\r
                passed = '✓' if r.get('passed', False) else '✗'\r
                overall = r.get('overall', r.get('recent_percentage', 'N/A'))\r
                if isinstance(overall, (int, float)):\r
                    report.append(f"{metric.upper():15s}: {overall:>6.1f}% {passed}")\r
                else:\r
                    report.append(f"{metric.upper():15s}: {overall}")\r
\r
        report.append("-" * 60)\r
\r
        if self.issues:\r
            report.append("")\r
            report.append("ISSUES FOUND:")\r
            for issue in self.issues[:10]:  # Show first 10\r
                report.append(f"  • {issue}")\r
            if len(self.issues) > 10:\r
                report.append(f"  ... and {len(self.issues) - 10} more issues")\r
\r
        report.append("")\r
        report.append("=" * 60)\r
\r
        return "\
".join(report)\r
```\r
\r
## Validation Rules Builder\r
\r
### Custom Validation Rules\r
\r
```python\r
class ValidationRulesBuilder:\r
    """Build custom validation rules for construction data"""\r
\r
    def __init__(self):\r
        self.rules = []\r
\r
    def add_not_null(self, column):\r
        """Column must not have null values"""\r
        self.rules.append({\r
            'type': 'not_null',\r
            'column': column,\r
            'check': lambda df, col=column: df[col].notna().all()\r
        })\r
        return self\r
\r
    def add_unique(self, column):\r
        """Column must have unique values"""\r
        self.rules.append({\r
            'type': 'unique',\r
            'column': column,\r
            'check': lambda df, col=column: df[col].nunique() == len(df)\r
        })\r
        return self\r
\r
    def add_range(self, column, min_val=None, max_val=None):\r
        """Column values must be within range"""\r
        self.rules.append({\r
            'type': 'range',\r
            'column': column,\r
            'min': min_val,\r
            'max': max_val,\r
            'check': lambda df, col=column, mn=min_val, mx=max_val:\r
                df[col].between(mn or -np.inf, mx or np.inf).all()\r
        })\r
        return self\r
\r
    def add_regex(self, column, pattern):\r
        """Column values must match regex pattern"""\r
        self.rules.append({\r
            'type': 'regex',\r
            'column': column,\r
            'pattern': pattern,\r
            'check': lambda df, col=column, p=pattern:\r
                df[col].astype(str).str.match(p).all()\r
        })\r
        return self\r
\r
    def add_in_list(self, column, valid_values):\r
        """Column values must be in list"""\r
        self.rules.append({\r
            'type': 'in_list',\r
            'column': column,\r
            'valid_values': valid_values,\r
            'check': lambda df, col=column, vals=valid_values:\r
                df[col].isin(vals).all()\r
        })\r
        return self\r
\r
    def add_custom(self, name, check_func):\r
        """Add custom validation function"""\r
        self.rules.append({\r
            'type': 'custom',\r
            'name': name,\r
            'check': check_func\r
        })\r
        return self\r
\r
    def validate(self, df):\r
        """Run all validation rules"""\r
        results = []\r
\r
        for rule in self.rules:\r
            try:\r
                passed = rule['check'](df)\r
                results.append({\r
                    'rule': rule.get('name', f"{rule['type']}:{rule.get('column', 'custom')}"),\r
                    'passed': passed,\r
                    'type': rule['type']\r
                })\r
            except Exception as e:\r
                results.append({\r
                    'rule': rule.get('name', f"{rule['type']}:{rule.get('column', 'custom')}"),\r
                    'passed': False,\r
                    'error': str(e)\r
                })\r
\r
        return results\r
\r
# Usage example\r
rules = (ValidationRulesBuilder()\r
    .add_not_null('ElementId')\r
    .add_unique('ElementId')\r
    .add_range('Volume_m3', min_val=0)\r
    .add_range('Cost', min_val=0)\r
    .add_in_list('Category', ['Wall', 'Floor', 'Column', 'Beam', 'Slab'])\r
    .add_regex('Level', r'^Level\s*\d+$')\r
)\r
\r
results = rules.validate(df)\r
for r in results:\r
    status = '✓' if r['passed'] else '✗'\r
    print(f"{status} {r['rule']}")\r
```\r
\r
## Automated Quality Pipeline\r
\r
```python\r
class DataQualityPipeline:\r
    """Automated data quality pipeline"""\r
\r
    def __init__(self, config=None):\r
        self.config = config or self._default_config()\r
        self.history = []\r
\r
    def _default_config(self):\r
        return {\r
            'required_columns': ['ElementId', 'Category', 'Volume_m3'],\r
            'unique_columns': ['ElementId'],\r
            'numeric_ranges': {\r
                'Volume_m3': (0, 10000),\r
                'Area_m2': (0, 100000),\r
                'Cost': (0, 100000000)\r
            },\r
            'valid_categories': ['Wall', 'Floor', 'Column', 'Beam', 'Slab',\r
                                 'Foundation', 'Roof', 'Stair', 'Door', 'Window'],\r
            'min_quality_score': 90\r
        }\r
\r
    def run(self, df, source_name='unknown'):\r
        """Run quality pipeline"""\r
        checker = DataQualityChecker(df)\r
\r
        # Configure checks based on config\r
        checker.check_completeness(self.config['required_columns'])\r
        checker.check_accuracy({\r
            col: {'min': r[0], 'max': r[1]}\r
            for col, r in self.config['numeric_ranges'].items()\r
        })\r
        checker.check_consistency(self.config['unique_columns'])\r
        checker.check_validity()\r
\r
        results = checker.run_full_check()\r
\r
        # Store in history\r
        self.history.append({\r
            'timestamp': datetime.now(),\r
            'source': source_name,\r
            'records': len(df),\r
            'score': results['overall_score'],\r
            'grade': results['grade'],\r
            'issues_count': len(results['issues'])\r
        })\r
\r
        # Check threshold\r
        passed = results['overall_score'] >= self.config['min_quality_score']\r
\r
        return {\r
            'passed': passed,\r
            'score': results['overall_score'],\r
            'grade': results['grade'],\r
            'details': results,\r
            'report': checker.generate_report()\r
        }\r
\r
    def get_history_summary(self):\r
        """Get quality history summary"""\r
        if not self.history:\r
            return "No quality checks performed yet."\r
\r
        df_history = pd.DataFrame(self.history)\r
        return {\r
            'total_checks': len(self.history),\r
            'avg_score': df_history['score'].mean(),\r
            'min_score': df_history['score'].min(),\r
            'max_score': df_history['score'].max(),\r
            'latest': self.history[-1]\r
        }\r
```\r
\r
## Quality Reporting\r
\r
### Export Quality Report\r
\r
```python\r
def export_quality_report(df, output_path, include_details=True):\r
    """Export comprehensive quality report to Excel"""\r
    checker = DataQualityChecker(df)\r
    results = checker.run_full_check()\r
\r
    with pd.ExcelWriter(output_path, engine='openpyxl') as writer:\r
        # Summary sheet\r
        summary = pd.DataFrame({\r
            'Metric': ['Overall Score', 'Grade', 'Records', 'Columns', 'Issues'],\r
            'Value': [\r
                f"{results['overall_score']:.1f}%",\r
                results['grade'],\r
                len(df),\r
                len(df.columns),\r
                len(results['issues'])\r
            ]\r
        })\r
        summary.to_excel(writer, sheet_name='Summary', index=False)\r
\r
        # Completeness details\r
        if 'completeness' in results:\r
            comp_df = pd.DataFrame.from_dict(\r
                results['completeness']['by_column'],\r
                orient='index',\r
                columns=['Completeness_%']\r
            )\r
            comp_df.to_excel(writer, sheet_name='Completeness')\r
\r
        # Issues list\r
        if results['issues']:\r
            issues_df = pd.DataFrame({'Issue': results['issues']})\r
            issues_df.to_excel(writer, sheet_name='Issues', index=False)\r
\r
        # Missing values analysis\r
        if include_details:\r
            missing = df.isnull().sum()\r
            missing_df = pd.DataFrame({\r
                'Column': missing.index,\r
                'Missing_Count': missing.values,\r
                'Missing_%': (missing.values / len(df) * 100).round(2)\r
            })\r
            missing_df.to_excel(writer, sheet_name='Missing_Values', index=False)\r
\r
    return output_path\r
```\r
\r
## Quick Reference\r
\r
| Metric | Description | Threshold |\r
|--------|-------------|-----------|\r
| Completeness | % non-null values | ≥ 95% |\r
| Accuracy | Values within valid range | ≥ 98% |\r
| Consistency | Unique IDs, valid relationships | ≥ 99% |\r
| Validity | Match expected patterns | ≥ 95% |\r
| Timeliness | Records updated recently | ≥ 80% |\r
\r
## Common Validation Patterns\r
\r
```python\r
# Construction-specific regex patterns\r
PATTERNS = {\r
    'element_id': r'^[A-Z]{1,3}\d{3,8}$',\r
    'revit_id': r'^\d{5,8}$',\r
    'ifc_guid': r'^[A-Za-z0-9_$]{22}$',\r
    'level': r'^(Level|L|Уровень)\s*[-]?\d+$',\r
    'grid': r'^[A-Z]{1,2}[-/]?\d{0,3}$',\r
    'date_iso': r'^\d{4}-\d{2}-\d{2}$',\r
    'cost_code': r'^\d{2,3}[.-]\d{2,4}[.-]?\d{0,4}$'\r
}\r
```\r
\r
## Resources\r
\r
- **Book**: "Data-Driven Construction" by Artem Boiko, Chapter 2.6\r
- **Website**: https://datadrivenconstruction.io\r
- **Great Expectations**: https://greatexpectations.io\r
\r
## Next Steps\r
\r
- See `bim-validation-pipeline` for BIM-specific validation\r
- See `etl-pipeline` for data processing pipelines\r
- See `data-visualization` for quality dashboards\r