Introduction
In today’s data-driven world, extracting meaningful insights from unstructured text is a major challenge—especially in fields like emergency services, where quick access to accurate information can save lives. My 2015 conference paper, The Cascading Knowledge Discovery in Databases (CKDD) Process in Information System Development, proposes a novel method to tackle this problem. By combining Knowledge Discovery in Databases (KDD) with qualitative analysis, this approach helps redesign information systems (IS) to better meet evolving user needs.
Key Takeaways
-
The Problem: Traditional IS often fail to adapt to new information requirements, leaving critical data buried in unstructured text (e.g., fire service reports).
-
The Solution: The CKDD process iteratively restructures data using:
-
Text Mining: Classifying and extracting patterns from natural language.
-
Failure Analysis: Tools like FMEA and SFTA to identify system weaknesses.
-
Formal Concept Analysis (FCA): Creating taxonomies for attributes (e.g., hydrant types, locations).
-
-
Case Study: Applied to Polish Fire Service reports, CKDD transformed unstructured narratives into a structured database of hydrant statuses, locations, and efficiency metrics.
How It Works
-
Step 1: Analyze existing IS (e.g., fire reports) to identify gaps using FMEA/SFTA.
-
Step 2: Use supervised learning to classify text segments (e.g., “operation,” “equipment”).
-
Step 3: Extract entities (e.g., hydrant IDs, street names) via pattern matching and FCA.
-
Step 4: Build a new IS with structured attributes, enabling faster, precise queries (e.g., “Show all efficient hydrants in District X”).
Results
-
73% of hydrants were operational; 2% were defective.
-
Geotagging revealed intervention hotspots (e.g., Warsaw’s Śródmieście district).
-
Semantic search improved accuracy by 40% compared to full-text queries.
Why It Matters
CKDD bridges the gap between static IS and dynamic real-world needs. It’s scalable for domains like healthcare, logistics, or public safety—anywhere unstructured data hides actionable insights.
Read the Full Paper: DOI Link or Research gate