Challenges and approaches for instrumenting and cleaning 'real'/ ugly data
Most practicing data scientists have those “bad data days” where you realize the data is corrupt, or not what you assumed the data to be, or labels are not right or even worse. What if we work in a paradigm assuming: “all data is corrupt, some is useful”, while at the same time instrumenting for any data which can be captured? In such a setting, how to go about various day-to-day data cleaning challenges, prioratizing data collection, building tooling for data quality etc?
Who should attend this BOF?
- Anyone who is eager to share stories around collecting richer data or making data cleaner in a systematic way
- Data geeks
- If you want to learn how to handle data more carefully and with respect
In this session we will share our experience on multiple things associated with collecting/instrumenting data and converting it to useful/accessible form. Specifically, the focus will be on emerging data in space of agriculture, biology, telemetry, images. Few of the topics which would be discussed are:
- Dealing with incomplete data
- Data accuracy issues
- Tools for data dictionary
- Process of data cleaning
- QA for right data instrumentation
We will cover the following topics in this session:
- Data basics like meta data store, data dictionary, documentation of data flow – challenges and tooling.
- Types of data issues including discovery of known unknowns: incomplete data, improper instrumentation, corrupted pipeline, varying quality, quality of labels – what are the reasons?
- Categorising sources and types of issues – ways of dealing with each type.
- Steps/milestones in journey for better data: legacy data, new data
- How to know what to instrument – various approaches.
- Suggestions and recommendations to approach the problem.
You should have played with data and felt the pain of something being off. :)
Kranthi Mitra – Principal Data Scientist at Swiggy
Raghotham Sripadraj – Senior Data Scientist at Ericsson
Elvis Joel D’Souza – Director of Product Engineering at Sensara
Karnam Vasudeva Rao – Senior Scientist in the data science team at Bayer