Data Science

How to handle data mining from multi-source heterogeneous databases without losing consistency?

KI Asked by Kimberly Sanders · 12-05-2023
0 upvotes 14,272 views 0 comments
The question

We’re struggling with a project involving data mining across several heterogeneous databases. The formats are inconsistent, and we’re seeing a lot of redundancy. What are the best strategies for integrating these multi-source streams while ensuring data integrity and minimizing cleaning time? 

3 answers

0
MA
Answered on 15-05-2023

Dealing with heterogeneous sources requires a robust ETL (Extract, Transform, Load) framework. In my experience at a retail giant, we used a "Schema Mapping" approach to create a unified view before mining. The key is to address "Object Equality"—identifying when records from different sources refer to the same entity. We automated this using fuzzy matching and Jaro-Winkler distance metrics. This significantly reduced our manual cleaning phase. Also, remember that data transformation like normalization is crucial here; if your scales aren't uniform, your clustering algorithms will produce skewed results. 

0
RI
Answered on 17-05-2023

Are you planning to use a centralized data warehouse for this, or are you looking into federated mining where the data stays in its original source during the analysis? 

ST 19-05-2023

Richard, we're actually leaning toward a data lakehouse architecture. It allows us to keep the raw data flexible while providing the structured querying needed for mining. To answer your question, we use metadata tagging at the point of ingestion. This way, even if the data stays "local" in a sense, the central mining engine knows exactly how to interpret the different schemas without having to physically move and transform every single byte first, saving us massive compute costs.

0
AM
Answered on 21-05-2023

You should focus heavily on the Data Purging phase. Removing NULL values and junk data early on will prevent your mining queries from becoming sluggish or inaccurate. 

K 22-05-2023

Exactly, Amanda. Garbage-in, garbage-out is the golden rule of data science. If the purging isn't strict, no amount of advanced algorithmic tuning will save the final insights.

Share your thoughts

Your email address will not be published. Required fields are marked (*)

Professional Counselling Session

Still have questions?
Schedule a free counselling session

Our experts are ready to help you with any questions about courses, admissions, or career paths. Get personalized guidance from industry professionals.

Request a Call Back

Search Online

We Accept

We Accept

Follow Us

"PMI®", "PMBOK®", "PMP®", "CAPM®" and "PMI-ACP®" are registered marks of the Project Management Institute, Inc. | "CSM", "CST" are Registered Trade Marks of The Scrum Alliance, USA. | COBIT® is a trademark of ISACA® registered in the United States and other countries.

Book Free Session