# API RESEARCH

## Daten von Isabella Molterer's Gruppe:
* https://docs.google.com/spreadsheets/d/1X7XhmFcIDHe3ejwkhRaaQACyVcnMszV-oNpa0PzCp7Y/edit#gid=336405600

* Erklärung für Seiten:
* Anbieter haben wir persönlich befragt und einen klick-prototyp testen lassen
* bei umfrage 1 siehst an wen wir allen einen Link zu unserer geschickt haben und wer danach interesse an unserem projekt gezeigt hat

## Links vom Flo

* LinkedIn:
* https://www.linkedin.com/in/oliver-wagner-85177a84/
* Kontant IMT:
* https://www.fh-salzburg.ac.at/suche?tx_solr%5Bq%5D=susanne+steppat

## timetable

* what is a fixture:
* rapid prototyping 21.-23.12.
* 12 january submission of project plan, project order, SUAS resources, final concept (topic, problems, solutions, ideas, sketches, ...)
* february studio week (1.-5.2.; one week of full time coding; mo - fr noon)
* friday afternoon of the studio week: user testing (users test what has been done - including brigitte)
* what did florian say - maybe confirm with brigitte
* rapid prototyping before christmas might not need to be coding
* we can even restart the whole project at the end of the second term
* we should formulate an email that could be sent to people in the tourism union

## notes
* what might be importantant to know?
* how many requests are free?
* how many requests are needed to get the required data?
* does it include data like location, opening hours, images, ...?
* does it include activities possible there?

## research

1. Datasets from data.salzburgerland - has in events and turist detsinations, unfortunatley from what I understood not updated since 3 yrs. Maybe make sens to contact SalzburgerLand Tourismus GmbH, and aks if they could provide us with sth...
[http://data.salzburgerland.com/dataset](http://data.salzburgerland.com/dataset) - FREE
[https://www.salzburg.info/](https://www.salzburg.info/)
[https://www.salzburgerland.com/](https://www.salzburgerland.com/)

2. Outdooractive - possible to get tours "nearby", some filtering options possible. Needs a project key[http://developers.outdooractive.com/API-Reference/Data-API.html](https://developers.outdooractive.com/API-Reference/Data-API.html) - FREE

3. Lonely planet API - in case someone didn't know, lonely planet are one of the most popular travel guides [https://docs.dev.content-api.lonelyplanet.com/#introduction](https://docs.dev.content-api.lonelyplanet.com/#introduction)

4. hiking and mountain biking trails, campgrounds, ski resorts, ATV trails, and more
https://rapidapi.com/trailapi/api/trailapi
5. Places by category API Documentation
https://rapidapi.com/geoapify-gmbh-geoapify/api/places-by-category
6. Google Places
* Request for every photo view (only with API key)
* either save it locally -> hugh storage needs
* or proxy every photo view -> hugh bandwidth and request count needs
* Can get all places in the vicinity in one request
* Only generic places - no activities
* one could use some business logic to fill the activities -> at a lake you can swim, ride a boat, ...
* one could use text analysis to determine what you might be able to do there -> intense research and feedback needed
7. TrueWay Places API Documentation
https://rapidapi.com/trueway/api/trueway-places
8. Geocoding Places
https://rapidapi.com/ai-box-ai-box-default/api/geocoding-places
9. Findery Api
https://rapidapi.com/findery/api/findery
With the Findery API, you can take advantage of our great content, near and far. Get nearby notes for local discovery, tourism, or the awesome tidbits you won’t find in history books. Display notes about your business to show how much customers love you. Create a notemap full of hidden treasures to visit. We look forward to finding new uses for Findery!
10.

## algorithms

* user based collaborative filtering - "items of similar users"
* pro
* "live" - new data/ratings can be used almost immediately
* con
* huge matix
* huge computation might be needed
* needs many many users initially
* item based collaborative filtering - "similar items"
* pro
* "live" - new data/ratings can be used almost immediately
* con
* huge matrix
* huge computation might be needed
* "filter bubble"
* incremental singular value decomposition (SVD)
* pro
* can be trained with incremental data -> faster than complete retrain -> can be done more often maybe even while running
* con
* huge matrix
* new training for each new user is needed
* xgboost with user id
* pro
* often good recommendations even based on features -> few recommendations can be compensated with others
* con
* needs to be trained completely for each new user and more or less new rating
* xgboost without user id
* pro
* often good recommendations even based on features -> few recommendations can be compensated with others
* can be used with just user features -> new users get results
* con
* needs to be retrained completely for new users/ratings
* does not really take user individuality into account