Innovation week dec12
Download as PPTX, PDF0 likes203 views
The document discusses how a valuations team extracts property data like number of beds and baths to estimate property prices using algorithms. Users can provide feedback on the estimates. The team's goals are to check for obvious mistakes in feedback, facilitate easy review of feedbacks, and potentially discover new useful features from feedback. Feedback is sorted based on detail, accuracy, source, and normalization. Text clustering is used to analyze feedback text and discover frequently occurring terms.
Innovation week dec12
- 2. What We Do Extract features from property data (# beds, # baths, sqft, past sales etc) Use a very complex algorithm to estimate price using these features Give users the option to register their feedback
- 3. What We Do Extract features from property data (# beds, # baths, sqft etc) Use a very complex algorithm to estimate price using these features Give users the option to register their feedback
- 6. Useless Junk
- 7. Goals Auto correction of the estimate based on the feedback? Checking for obvious mistakes? Facilitating easy review of the feedbacks? Discovering new useful features?
- 8. Goals Auto correction of the estimate based on the feedback? No Checking for obvious mistakes? Facilitating easy review of the feedbacks? Discovering new useful features?
- 9. Goals Auto correction of the estimate based on the feedback? Checking for obvious mistakes? Yes Facilitating easy review of the feedbacks? Discovering new useful features?
- 10. Goals Auto correction of the estimate based on the feedback? Checking for obvious mistakes? Facilitating easy review of the feedbacks? Yes Discovering new useful features?
- 11. Goals Auto correction of the estimate based on the feedback? Checking for obvious mistakes? Facilitating easy review of the feedbacks? Discovering new useful features? Perhaps
- 12. Unsorted Feedback http://analyticsnn2.sv2.trulia.com:8111/feedback-2012-05-21.html
- 13. After Sorting
- 14. After Sorting
- 15. What is the Sorting Based on? How detailed is the replay? How far off are we? Did a registered user submit the feedback? Normalization/ regularization
- 16. Distribution by Geographical Scope Global Average: 1.2178
- 17. Text Clustering Use Apache Mahout to cluster feedback text Discover terms that most frequently occur together in feedback texts