open-construction-estimate

Open Construction Estimate

Overview

This skill leverages open construction pricing databases for automated cost estimation. Match project elements to standardized work items and calculate costs using publicly available unit prices.

Data Sources:

• OpenConstructionEstimate (55,000+ work items)
• RSMeans Online (subscription)
• Government pricing databases
• Regional cost indexes

"Открытые базы данных расценок содержат более 55,000 позиций работ, что позволяет автоматизировать сметные расчеты для большинства проектов."

— DDC LinkedIn

Quick Start

import pandas as pd
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.metrics.pairwise import cosine_similarity

# Load work items database
work_items = pd.read_csv("open_construction_estimate.csv")
print(f"Loaded {len(work_items)} work items")

# Simple matching function
vectorizer = TfidfVectorizer(ngram_range=(1, 2))
item_vectors = vectorizer.fit_transform(work_items['description'])

def find_matching_items(query, top_n=5):
    query_vec = vectorizer.transform([query])
    similarities = cosine_similarity(query_vec, item_vectors)[0]
    top_indices = similarities.argsort()[-top_n:][::-1]

    return work_items.iloc[top_indices][['code', 'description', 'unit', 'unit_price']]

# Find matches
matches = find_matching_items("reinforced concrete wall 300mm")
print(matches)

Open Database Structure

Database Schema

# Standard work items database structure
WORK_ITEMS_SCHEMA = {
    'code': 'Work item code (e.g., 03.31.13.13)',
    'description': 'Full description of work',
    'short_description': 'Abbreviated description',
    'unit': 'Unit of measure (m³, m², ton, pcs)',
    'unit_price': 'Base unit price',
    'labor_cost': 'Labor component per unit',
    'material_cost': 'Material component per unit',
    'equipment_cost': 'Equipment component per unit',
    'labor_hours': 'Labor hours per unit',
    'crew_size': 'Typical crew size',
    'productivity': 'Units per day',
    'category_l1': 'Primary category (CSI Division)',
    'category_l2': 'Secondary category',
    'category_l3': 'Detailed category',
    'region': 'Geographic region',
    'year': 'Price year',
    'source': 'Data source'
}

# CSI MasterFormat Divisions
CSI_DIVISIONS = {
    '03': 'Concrete',
    '04': 'Masonry',
    '05': 'Metals',
    '06': 'Wood, Plastics, Composites',
    '07': 'Thermal and Moisture Protection',
    '08': 'Openings',
    '09': 'Finishes',
    '10': 'Specialties',
    '21': 'Fire Suppression',
    '22': 'Plumbing',
    '23': 'HVAC',
    '26': 'Electrical',
    '31': 'Earthwork',
    '32': 'Exterior Improvements',
    '33': 'Utilities'
}

Work Item Matching Engine

Semantic Matching System

import pandas as pd
import numpy as np
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.metrics.pairwise import cosine_similarity
from sentence_transformers import SentenceTransformer
from typing import List, Dict, Optional, Tuple
import re

class WorkItemMatcher:
    """Match BIM elements to standardized work items"""

    def __init__(self, database_path: str, use_embeddings: bool = True):
        self.db = pd.read_csv(database_path)

        # TF-IDF for fast initial filtering
        self.tfidf = TfidfVectorizer(
            ngram_range=(1, 3),
            max_features=10000,
            stop_words='english'
        )
        self.tfidf_matrix = self.tfidf.fit_transform(self.db['description'])

        # Sentence embeddings for semantic matching
        self.use_embeddings = use_embeddings
        if use_embeddings:
            self.embedder = SentenceTransformer('all-MiniLM-L6-v2')
            self.embeddings = self.embedder.encode(
                self.db['description'].tolist(),
                show_progress_bar=True
            )

    def match(self, query: str, top_n: int = 5,
              category: str = None) -> List[Dict]:
        """Find matching work items for a query"""
        # Filter by category if specified
        if category:
            mask = self.db['category_l1'].str.contains(category, case=False, na=False)
            search_db = self.db[mask]
            search_matrix = self.tfidf_matrix[mask]
        else:
            search_db = self.db
            search_matrix = self.tfidf_matrix

        if self.use_embeddings:
            return self._semantic_match(query, search_db, top_n)
        else:
            return self._tfidf_match(query, search_db, search_matrix, top_n)

    def _tfidf_match(self, query: str, db: pd.DataFrame,
                     matrix, top_n: int) -> List[Dict]:
        """TF-IDF based matching"""
        query_vec = self.tfidf.transform([query])
        similarities = cosine_similarity(query_vec, matrix)[0]

        top_indices = similarities.argsort()[-top_n:][::-1]

        results = []
        for idx in top_indices:
            row = db.iloc[idx]
            results.append({
                'code': row['code'],
                'description': row['description'],
                'unit': row['unit'],
                'unit_price': row['unit_price'],
                'similarity': float(similarities[idx]),
                'category': row.get('category_l1', '')
            })

        return results

    def _semantic_match(self, query: str, db: pd.DataFrame,
                        top_n: int) -> List[Dict]:
        """Semantic embedding based matching"""
        query_embedding = self.embedder.encode([query])

        # Get indices for filtered db
        indices = db.index.tolist()
        filtered_embeddings = self.embeddings[indices]

        similarities = cosine_similarity(query_embedding, filtered_embeddings)[0]
        top_indices = similarities.argsort()[-top_n:][::-1]

        results = []
        for i, idx in enumerate(top_indices):
            row = db.iloc[idx]
            results.append({
                'code': row['code'],
                'description': row['description'],
                'unit': row['unit'],
                'unit_price': row['unit_price'],
                'similarity': float(similarities[idx]),
                'category': row.get('category_l1', '')
            })

        return results

    def match_bim_element(self, element: Dict) -> List[Dict]:
        """Match a BIM element to work items"""
        # Build query from element properties
        query_parts = []

        if element.get('material'):
            query_parts.append(element['material'])
        if element.get('category'):
            query_parts.append(element['category'])
        if element.get('description'):
            query_parts.append(element['description'])

        # Add dimensions if available
        if element.get('thickness'):
            query_parts.append(f"{element['thickness']}mm thick")
        if element.get('height'):
            query_parts.append(f"{element['height']}m high")

        query = ' '.join(query_parts)

        # Determine category from element type
        category = self._get_category_from_element(element)

        return self.match(query, top_n=3, category=category)

    def _get_category_from_element(self, element: Dict) -> Optional[str]:
        """Map BIM element type to CSI category"""
        element_mapping = {
            'IfcWall': 'Concrete|Masonry',
            'IfcSlab': 'Concrete',
            'IfcColumn': 'Concrete|Metals',
            'IfcBeam': 'Concrete|Metals',
            'IfcDoor': 'Openings',
            'IfcWindow': 'Openings',
            'IfcRoof': 'Thermal',
            'IfcStair': 'Concrete',
            'IfcPipeSegment': 'Plumbing',
            'IfcDuctSegment': 'HVAC'
        }

        elem_type = element.get('ifc_type', '')
        return element_mapping.get(elem_type)

Cost Estimation Engine

Automated Estimator

class OpenConstructionEstimator:
    """Generate cost estimates using open databases"""

    def __init__(self, matcher: WorkItemMatcher, region: str = 'default'):
        self