Insight Project Demo

This is a demo of an interactive platform for data analytics. This 3-week project was carried out as part of the Insight Data Science Fellowship.

Transcript

1. Ikenna Dominic Ivenso Fellow, Insight Data Science (Summer 2016) Interac>ve Pla@orm for Data Analy>cs
2. Problem: •  Opera>ng en>>es collect large amounts of data •  Mostly unexplored due to lack of exper>se Solu*on: Develop pla@orm to help users… •  Gain insights from data •  Improve opera>ons
3. Example: Team of Healthcare Workers on Outreach Program Tasks: •  Test pa*ents for factors that predict onset of diabetes •  And provide *mely diagnosis of diabetes Available: •  Test results for several pa*ents •  Access to personal computer
4. Preprocessing: * Data Extrac>on * One-Hot-Encoding/ Binarizing * Feature Scaling
5. Preprocessing: * Data Extrac>on * One-Hot-Encoding/ Binarizing * Feature Scaling Feature Selec>on: * Recursive Feature Elimina>on * Stra>fied K-Fold Cross-Valida>on * L1 Regulariza>on (for Correla>on and Sparsity)
6. Preprocessing: * Data Extrac>on * One-Hot-Encoding/ Binarizing * Feature Scaling Feature Selec>on: * Recursive Feature Elimina>on * Stra>fied K-Fold Cross-Valida>on * L1 Regulariza>on (for Correla>on and Sparsity) Feature Importance Ranking: * Random Forest Classifier
7. Preprocessing: * Data Extrac>on * One-Hot-Encoding/ Binarizing * Feature Scaling Feature Selec>on: * Recursive Feature Elimina>on * Stra>fied K-Fold Cross-Valida>on * L1 Regulariza>on (for Correla>on and Sparsity) Feature Importance Ranking: * Random Forest Classifier Report: * Matplotlib, Bokeh
8. Summary PlaBorm gives feedback to user on… •  Important features of their data •  Actual impact of each feature PlaBorm is flexible •  Handles categorical and numeric data •  Applicable to wide variety of data sources The plaBorm can be accessed at: •  hGp://www.ikennaivenso.xyz/upload
9. Ph.D. Mechanical Engineering, Texas Tech University Ikenna Ivenso Built computa>onal models of DNA HUUUGE soccer fan!!!