historical-data-manager

建設分野の過去データ管理

概要

アーカイブ、旧システム、過去の記録から建設分野のレガシーデータを管理します。データを抽出、クリーンアップ、正規化し、分析やベンチマークに使える最新の形式へ移行します。

ビジネス上のメリット

建設会社には、さまざまな形式で数十年分のプロジェクトデータが蓄積されています。

• PDF にスキャンされた紙の記録
• レガシーデータベースのエクスポート（Access、dBase、FoxPro）
• 古いスプレッドシート形式（Lotus 1-2-3、初期の Excel）
• 独自ソフトウェアからのエクスポート
• プロジェクト完了時の文書

このスキルは、過去のデータを活用して以下を実現します。

• コストのベンチマークと推移分析
• 経時的な生産性分析
• リスクパターンの特定
• 見積もりの改善

技術的な実装

過去データの抽出ツール

from dataclasses import dataclass, field
from typing import List, Dict, Any, Optional
from datetime import datetime
from pathlib import Path
import pandas as pd
import re
import json

@dataclass
class HistoricalRecord:
    project_id: str
    project_name: str
    year: int
    data_type: str  # cost, schedule, labor, material
    original_format: str
    extracted_data: Dict[str, Any]
    quality_score: float
    notes: List[str] = field(default_factory=list)

class HistoricalDataManager:
    """Manage extraction and normalization of historical construction data."""

    def __init__(self, archive_path: str):
        self.archive_path = Path(archive_path)
        self.records: List[HistoricalRecord] = []
        self.normalization_rules = self._load_normalization_rules()

    def scan_archive(self) -> Dict[str, int]:
        """Scan archive and categorize files by type."""
        file_types = {}

        for file_path in self.archive_path.rglob('*'):
            if file_path.is_file():
                ext = file_path.suffix.lower()
                file_types[ext] = file_types.get(ext, 0) + 1

        return file_types

    def extract_from_legacy_excel(self, file_path: str, year: int) -> List[HistoricalRecord]:
        """Extract data from legacy Excel files."""
        records = []

        try:
            # Try different engines for old formats
            try:
                df = pd.read_excel(file_path, engine='openpyxl')
            except:
                df = pd.read_excel(file_path, engine='xlrd')

            # Detect data type from content
            data_type = self._detect_data_type(df)

            # Normalize column names
            df = self._normalize_columns(df)

            # Extract project info
            project_info = self._extract_project_info(df, file_path)

            record = HistoricalRecord(
                project_id=project_info.get('id', f'LEGACY-{year}-{hash(file_path) % 10000}'),
                project_name=project_info.get('name', Path(file_path).stem),
                year=year,
                data_type=data_type,
                original_format='excel',
                extracted_data=df.to_dict('records'),
                quality_score=self._assess_quality(df)
            )
            records.append(record)

        except Exception as e:
            print(f"Error extracting {file_path}: {e}")

        return records

    def extract_from_csv(self, file_path: str, year: int) -> HistoricalRecord:
        """Extract data from CSV files with encoding detection."""
        # Try different encodings
        encodings = ['utf-8', 'latin-1', 'cp1252', 'iso-8859-1']

        for encoding in encodings:
            try:
                df = pd.read_csv(file_path, encoding=encoding)
                break
            except:
                continue

        df = self._normalize_columns(df)
        data_type = self._detect_data_type(df)

        return HistoricalRecord(
            project_id=f'CSV-{year}-{hash(file_path) % 10000}',
            project_name=Path(file_path).stem,
            year=year,
            data_type=data_type,
            original_format='csv',
            extracted_data=df.to_dict('records'),
            quality_score=self._assess_quality(df)
        )

    def extract_from_database_export(self, file_path: str, db_type: str) -> List[HistoricalRecord]:
        """Extract data from legacy database exports."""
        records = []

        if db_type == 'access':
            # Read Access MDB/ACCDB files
            import pyodbc
            conn_str = f'DRIVER={{Microsoft Access Driver (*.mdb, *.accdb)}};DBQ={file_path}'
            conn = pyodbc.connect(conn_str)

            # Get all tables
            cursor = conn.cursor()
            tables = [row.table_name for row in cursor.tables(tableType='TABLE')]

            for table in tables:
                df = pd.read_sql(f'SELECT * FROM [{table}]', conn)
                # Process each table...

            conn.close()

        return records

    def normalize_cost_data(self, records: List[HistoricalRecord], base_year: int = 2026) -> pd.DataFrame:
        """Normalize historical cost data to current dollars."""
        # RSMeans historical cost indices (example values)
        cost_indices = {
            2015: 0.82, 2016: 0.84, 2017: 0.87, 2018: 0.90,
            2019: 0.93, 2020: 0.95, 2021: 0.98, 2022: 1.02,
            2023: 1.06, 2024: 1.10, 2025: 1.14, 2026: 1.18
        }

        normalized_data = []

        for record in records:
            if record.data_type == 'cost':
                year_index = cost_indices.get(record.year, 1.0)
                base_index = cost_indices.get(base_year, 1.18)
                escalation_factor = base_index / year_index

                for item in record.extracted_data:
                    if 'amount' in item or 'cost' in item:
                        original_cost = item.get('amount') or item.get('cost', 0)
                        normalized_item = item.copy()
                        normalized_item['original_cost'] = original_cost
                        normalized_item['normalized_cost'] = original_cost * escalation_factor
                        normalized_item['escalation_factor'] = escalation_factor
                        normalized_item['original_year'] = record.year
                        normalized_item['project_id'] = record.project_id
                        normalized_data.append(normalized_item)

        return pd.DataFrame(normalized_data)

    def _detect_data_type(self, df: pd.DataFrame) -> str:
        """Detect type of data from column names and content."""
        columns_lower = [c.lower() for c in df.columns]

        if any(c in columns_lower for c in ['cost', 'amount', 'price', 'total', 'budget']):
            return 'cost'
        elif any(c in columns_lower for c in ['start', 'finish', 'duration', 'task', 'activity']):
            return 'schedule'
        elif any(c in columns_lower for c in ['hours', 'labor', 'worker', 'crew']):
            return 'labor'
        elif any(c in columns_lower for c in ['material', 'quantity', 'unit', 'supplier']):
            return 'material'
        else:
            return 'unknown'

    def _normalize_columns(self, df: pd.DataFrame) -> pd.DataFrame:
        """Normalize column names to standard format."""
        column_mapping = {
            r'proj.*id': 'project_id',
            r'proj.*name': 'project_name',
            r'desc.*': 'description',
            r'qty|quantity': 'quantity',
            r'unit.*cost|unit.*price': 'unit_cost',
            r'total|amount': 'amount',
            r'start.*date': 'start_date',
            r'end.*date|finish.*date': 'end_date',
            r'dur.*': 'duration',
        }

        new_columns = {}
        for col in df.columns:
            col_lower = col.lower().strip()
            for pattern, new_name in column_mapping.items():
                if re.match(pattern, col_lower):
                    new_columns[col] = new_name
                    break

        return df.rename(columns=new_columns)

    def _assess_quality(self, df: pd.DataFrame) -> float:
        """Assess data quality score (0-1)."""
        if df.empty:
            return 0.0

        scores = []

        # Completeness: % of non-null values
        completeness = 1 - (df.isnull().sum().sum() / df.size)
        scores.append(completeness)

        # Column quality: has meaningful column names
        meaningful_cols = sum(1 for c in df.columns if len(c) > 2 and not c.startswith('Unnamed'))
        col_quality = meaningful_cols / len(df.columns)
        scores.append(col_quality)

        # Row count: more data is better (capped at 1.0)
        row_score = min(len(df) / 100, 1.0)
        scores.append(row_score)

        return sum(scores) / len(scores)

    def _extract_project_info(self, df: pd.DataFrame, file_path: str) -> Dict[str, str]:
        """Extract project info from data or filename."""
        info = {}

        # Try to find project info in data
        for col in df.columns:
            if 'project' in col.lower() and 'id' in col.lower():
                info['id'] = str(df[col].iloc[0]) if not df[col].empty else None
            if 'project' in col.lower() and 'name' in col.lower():
                info['name'] = str(df[col].iloc[0]) if not df[col].empty else None

        # Fallback to filename
        if 'name' not in info:
            info['name'] = Path(file_path).stem

        return info

    def _load_normalization_rules(self) -> Dict:
        """Load rules for normalizing legacy data."""
        return {
            'unit_conversions': {
                'M': 1000,  # Thousand
                'C': 100,   # Hundred
                'LF': 1,    # Linear Foot
                'SF': 1,    # Square Foot
                'CY': 1,    # Cubic Yard
            },
            'date_formats': [
                '%m/%d/%Y', '%m/%d/%y', '%Y-%m-%d',
                '%d-%b-%Y', '%B %d, %Y'
            ]
        }

    def generate_migration_report(self) -> str:
        """Generate report on migrated data."""
        report = ["# Historical Data Migration Report", ""]

        # Summary
        report.append("## Summary")
        report.append(f"- Total Records: {len(self.records)}")

        by_type = {}
        by_year = {}
        for r in self.records:
            by_type[r.data_type] = by_type.get(r.data_type, 0) + 1
            by_year[r.year] = by_year.get(r.year, 0) + 1

        report.append("\n### By Data Type")
        for dt, count in sorted(by_type.items()):
            report.append(f"- {dt}: {count}")

        report.append("\n### By Year")
        for year, count in sorted(by_year.items()):
            report.append(f"- {year}: {count}")

        # Quality Assessment
        report.append("\n## Data Quality")
        avg_quality = sum(r.quality_score for r in self.records) / len(self.records) if self.records else 0
        report.append(f"- Average Quality Score: {avg_quality:.2%}")

        low_quality = [r for r in self.records if r.quality_score < 0.5]
        if low_quality:
            report.append(f"\n### Low Quality Records ({len(low_quality)})")
            for r in low_quality[:10]:
                report.append(f"- {r.project_name} ({r.year}): {r.quality_score:.2%}")

        return "\n".join(report)

レガシーシステム用コネクター

class LegacySystemConnector:
    """Connect to various legacy construction systems."""

    @staticmethod
    def read_timberline_export(file_path: str) -> pd.DataFrame:
        """Read Sage Timberline (now Sage 300) export files."""
        # Timberline exports typically have specific format
        df = pd.read_csv(file_path, encoding='cp1252')

        # Map Timberline columns to standard
        column_map = {
            'JOB': 'project_id',
            'PHASE': 'phase_code',
            'CATEGORY': 'cost_code',
            'DESCRIPTION': 'description',
            'ESTIMATE': 'estimated_cost',
            'ACTUAL': 'actual_cost',
            'COMMITTED': 'committed_cost'
        }

        return df.rename(columns=column_map)

    @staticmethod
    def read_primavera_xer(file_path: str) -> Dict[str, pd.DataFrame]:
        """Read Primavera P6 XER export files."""
        tables = {}
        current_table = None
        current_data = []
        columns = []

        with open(file_path, 'r', encoding='utf-8') as f:
            for line in f:
                line = line.strip()
                if line.startswith('%T'):
                    # Save previous table
                    if current_table and current_data:
                        tables[current_table] = pd.DataFrame(current_data, columns=columns)
                    # Start new table
                    current_table = line.split('\t')[1] if '\t' in line else None
                    current_data = []
                    columns = []
                elif line.startswith('%F'):
                    # Field definitions
                    columns = line.split('\t')[1:]
                elif line.startswith('%R'):
                    # Data row
                    current_data.append(line.split('\t')[1:])

        # Save last table
        if current_table and current_data:
            tables[current_table] = pd.DataFrame(current_data, columns=columns)

        return tables

    @staticmethod
    def read_mc2_ice(file_path: str) -> pd.DataFrame:
        """Read MC2 ICE estimating export."""
        # MC2 ICE format handling
        pass

クイックスタート

# Initialize manager
manager = HistoricalDataManager('/archive/projects')

# Scan archive
file_types = manager.scan_archive()
print(f"Found: {file_types}")

# Extract from legacy Excel files
for year in range(2015, 2024):
    year_path = f'/archive/projects/{year}'
    for file in Path(year_path).glob('*.xls*'):
        records = manager.extract_from_legacy_excel(str(file), year)
        manager.records.extend(records)

# Normalize cost data to 2026 dollars
cost_records = [r for r in manager.records if r.data_type == 'cost']
normalized_costs = manager.normalize_cost_data(cost_records, base_year=2026)

# Generate migration report
report = manager.generate_migration_report()
print(report)

# Export for analysis
normalized_costs.to_excel('historical_costs_normalized.xlsx', index=False)

主なユースケース

1. コストのベンチマーク：過去のコストを比較できるように正規化します
2. 生産性分析：労働生産性を経時的に追跡します
3. リスクの特定：過去のプロジェクトで発生した問題のパターンを見つけます
4. 見積もりの調整：過去のデータを活用して見積もりを改善します

依存関係

pip install pandas openpyxl xlrd pyodbc

リソース

• RSMeans Historical Cost Index：コストのエスカレーションに使用します
• ENR Construction Cost Index：代替のエスカレーション情報源
• レガシー形式のドキュメント：ベンダー固有のエクスポート形式