tweather
TRANSCRIPT
![Page 1: tweather](https://reader033.vdocument.in/reader033/viewer/2022052900/5561a2e5d8b42afd708b4a33/html5/thumbnails/1.jpg)
THE SITUATION
![Page 2: tweather](https://reader033.vdocument.in/reader033/viewer/2022052900/5561a2e5d8b42afd708b4a33/html5/thumbnails/2.jpg)
Everyone wants current local weather
but weather apps are often WRONG
![Page 3: tweather](https://reader033.vdocument.in/reader033/viewer/2022052900/5561a2e5d8b42afd708b4a33/html5/thumbnails/3.jpg)
![Page 4: tweather](https://reader033.vdocument.in/reader033/viewer/2022052900/5561a2e5d8b42afd708b4a33/html5/thumbnails/4.jpg)
![Page 5: tweather](https://reader033.vdocument.in/reader033/viewer/2022052900/5561a2e5d8b42afd708b4a33/html5/thumbnails/5.jpg)
![Page 6: tweather](https://reader033.vdocument.in/reader033/viewer/2022052900/5561a2e5d8b42afd708b4a33/html5/thumbnails/6.jpg)
BAYESIAN FILTERINGstep 1 develop keyword search and assemble
training set (bag of words)
step 2 construct Naive Bayes filter with training set
rinse and repeat
step 3 keyword search and pass through NB filter
![Page 7: tweather](https://reader033.vdocument.in/reader033/viewer/2022052900/5561a2e5d8b42afd708b4a33/html5/thumbnails/7.jpg)
RESULTS
accuracy of ~ 80-90%
cross validation > 95%
![Page 8: tweather](https://reader033.vdocument.in/reader033/viewer/2022052900/5561a2e5d8b42afd708b4a33/html5/thumbnails/8.jpg)
FUNDAMENTALS
rainoutside
cooliphone
0.960.68
0.410.55
lol 0.34
probability
![Page 9: tweather](https://reader033.vdocument.in/reader033/viewer/2022052900/5561a2e5d8b42afd708b4a33/html5/thumbnails/9.jpg)
FUNDAMENTALS
“it’s cool outside” 0.65
0.47
0.50
“my iphone is cool”
“it’s cool outside. will charge my iphone...lol!”
probability
![Page 10: tweather](https://reader033.vdocument.in/reader033/viewer/2022052900/5561a2e5d8b42afd708b4a33/html5/thumbnails/10.jpg)
PROCESS
LIVE tweets
![Page 11: tweather](https://reader033.vdocument.in/reader033/viewer/2022052900/5561a2e5d8b42afd708b4a33/html5/thumbnails/11.jpg)
ARISTOTLE SOCRATES