際際滷

際際滷Share a Scribd company logo

MSK_Presentation

?
?
N
Noah Kiwanuka

This document describes VacAdvisor, a tool that recommends vacation options based on a user's specified budget. It clusters over 720 US cities using data on flight costs, hotel rates, daily expenses, and location attributes. The conceptual framework involves clustering algorithms like k-means to group similar cities. Validation tests various numbers of clusters and algorithms, with k-means providing the best results based on metrics like within-sum-of-squares and adjusted rand index. The goal is to match users to vacation spots optimally based on their preferences and budgets.

1 of 14
Download to read offline
VacAdvisor Fred N. Kiwanuka Fellow Insight Data Science July 21, 2015 Fred N. Kiwanuka Fellow Insight Data Science VacAdvisor
What are my options for vacation with a speci?ed budget ? Fred N. Kiwanuka Fellow Insight Data Science VacAdvisor
The Data Fred N. Kiwanuka Fellow Insight Data Science VacAdvisor
The Data Fred N. Kiwanuka Fellow Insight Data Science VacAdvisor
The Data Fred N. Kiwanuka Fellow Insight Data Science VacAdvisor
The Data Fred N. Kiwanuka Fellow Insight Data Science VacAdvisor
Features Flight cost from New York to 720 cities in the US Average cost of hotels for 3143 counties Average daily expenses including car Location of city [east, west, central, south, north] City speci?cs like beaches, museum, national parks Fred N. Kiwanuka Fellow Insight Data Science VacAdvisor
Conceptual Framework Fred N. Kiwanuka Fellow Insight Data Science VacAdvisor
Algorithm: Clustering Fred N. Kiwanuka Fellow Insight Data Science VacAdvisor
Algorithm: Clustering Fred N. Kiwanuka Fellow Insight Data Science VacAdvisor
Algorithm: Clustering Fred N. Kiwanuka Fellow Insight Data Science VacAdvisor
Cluster Validation Table: (Cluster Validation) Number of Clusters WSS(105) City Similarity 2 9.98 Seattle Detroit, Charlotte, South Bend 3 9.73 Seattle Boston, Phoenix, Detroit 4 9.87 Seattle Charlotte, South Bend, Minneapolis 5 9.15 Seattle Detroit, Charlotte, South Bend 7 9.43 Seattle Detroit, Charlotte, South Bend 10 9.63 Seattle Sacrameto, San Jose, Colombus 12 9.52 Seattle Sacrameto, San Jose, Colombus Fred N. Kiwanuka Fellow Insight Data Science VacAdvisor
Cluster Initialization and Validation Table: (Cluster Initialization and Validation) Alg Time(s) homo compl v-meas ARI AMI silhouette k-means 0.03 0.971 0.971 0.971 0.988 0.970 0.389 VQ 0.04 1.000 1.000 1.000 1.000 1.000 0.388 After PCA 0.00 1.000 1.000 1.000 1.000 1.000 0.388 Mean Shift 0.24 1.000 0.970 0.972 0.980 0.972 0.386 Fred N. Kiwanuka Fellow Insight Data Science VacAdvisor
Fred N. Kiwanuka PhD(Groningen), MSC(London), MIT(Teaching Fellow) Fred N. Kiwanuka Fellow Insight Data Science VacAdvisor

More Related Content

MSK_Presentation

  • 1. VacAdvisor Fred N. Kiwanuka Fellow Insight Data Science July 21, 2015 Fred N. Kiwanuka Fellow Insight Data Science VacAdvisor
  • 2. What are my options for vacation with a speci?ed budget ? Fred N. Kiwanuka Fellow Insight Data Science VacAdvisor
  • 3. The Data Fred N. Kiwanuka Fellow Insight Data Science VacAdvisor
  • 4. The Data Fred N. Kiwanuka Fellow Insight Data Science VacAdvisor
  • 5. The Data Fred N. Kiwanuka Fellow Insight Data Science VacAdvisor
  • 6. The Data Fred N. Kiwanuka Fellow Insight Data Science VacAdvisor
  • 7. Features Flight cost from New York to 720 cities in the US Average cost of hotels for 3143 counties Average daily expenses including car Location of city [east, west, central, south, north] City speci?cs like beaches, museum, national parks Fred N. Kiwanuka Fellow Insight Data Science VacAdvisor
  • 8. Conceptual Framework Fred N. Kiwanuka Fellow Insight Data Science VacAdvisor
  • 9. Algorithm: Clustering Fred N. Kiwanuka Fellow Insight Data Science VacAdvisor
  • 10. Algorithm: Clustering Fred N. Kiwanuka Fellow Insight Data Science VacAdvisor
  • 11. Algorithm: Clustering Fred N. Kiwanuka Fellow Insight Data Science VacAdvisor
  • 12. Cluster Validation Table: (Cluster Validation) Number of Clusters WSS(105) City Similarity 2 9.98 Seattle Detroit, Charlotte, South Bend 3 9.73 Seattle Boston, Phoenix, Detroit 4 9.87 Seattle Charlotte, South Bend, Minneapolis 5 9.15 Seattle Detroit, Charlotte, South Bend 7 9.43 Seattle Detroit, Charlotte, South Bend 10 9.63 Seattle Sacrameto, San Jose, Colombus 12 9.52 Seattle Sacrameto, San Jose, Colombus Fred N. Kiwanuka Fellow Insight Data Science VacAdvisor
  • 13. Cluster Initialization and Validation Table: (Cluster Initialization and Validation) Alg Time(s) homo compl v-meas ARI AMI silhouette k-means 0.03 0.971 0.971 0.971 0.988 0.970 0.389 VQ 0.04 1.000 1.000 1.000 1.000 1.000 0.388 After PCA 0.00 1.000 1.000 1.000 1.000 1.000 0.388 Mean Shift 0.24 1.000 0.970 0.972 0.980 0.972 0.386 Fred N. Kiwanuka Fellow Insight Data Science VacAdvisor
  • 14. Fred N. Kiwanuka PhD(Groningen), MSC(London), MIT(Teaching Fellow) Fred N. Kiwanuka Fellow Insight Data Science VacAdvisor