Kickstarter Data Project

Slide 1

Slide 1 text

KICKSTArTER SUCCEss

Slide 2

Slide 2 text

Our Data. We Had 87000 projects. 300,000 users. We Used 15000 projects 400,000 users (with repetitions). We Collected Twitter + Facebook Shares. Project Creator’s Facebook Friend Count.

Slide 3

Slide 3 text

Cleansing. We Were Left With 12000 projects. We Chose Projects with >1 backers. Log scale. We Discarded Facebook data (extremely inconsistent)

Slide 4

Slide 4 text

Hypothesis. Global reach aﬀects the success of projects.

Slide 5

Slide 5 text

Backers. Hexbin Density Plot

Slide 6

Slide 6 text

Projects. Hexbin Density Plot

Slide 7

Slide 7 text

Features. Backer Location Variance Presence of Video Backers Comments Images Pledges Minimum Pledge Maximum Pledge Tweets Facebook Friends Art Dance Design Fashion Film/Video Food Games Music Publishing Technology

Slide 8

Slide 8 text

Backers Comments Pledges Updates

Slide 9

Slide 9 text

No content

Slide 10

Slide 10 text

Signiﬁcant Features. Backer Location Variance 2e-16 Has video 0.000455 Backers 2e-16 Comments 2e-16 Image Count 0.000290 Maximum Pledge 2e-16 Tweet Count 3.75e-05 Design 3.60e-05 Games 3.25e-14 Technology 0.001938

Slide 11

Slide 11 text

(without Backers) Comics 0.049473 Design 3.60e-05 Food 0.013668 Games 3.25e-14 Technology 0.001938 Backer Location Variance 0.001512 Comments 2e-16 Image Count 0.000524 Number of Pledges 0.017430 Maximum Pledge 2e-16 Tweet Count 9.14e-06 Signiﬁcant Features.

Slide 12

Slide 12 text

1504 130 78 3033 Classiﬁers. Random Forest 94% Accuracy

Slide 13

Slide 13 text

Classiﬁers. 0 0 1634 3111 3052 1516 59 118 GaussianRBF 65% Accuracy Linear 96.3% Accuracy SVM

Slide 14

Slide 14 text

Density plot: Hidden Radio and Bluetooth Speaker