Introduction

PetFinder.my Adoption Prediction is a Kaggle challenge where the goal is to predict the adoptability of pets using their online profiles and metadata provided by PetFinder.my, Malaysia’s leading animal welfare platform.

Supervisor: Erwan Scornet (École Polytechnique)

Data

Training data: 14993 pets
Testing data: 3972 pets

Data Fields

Structured Data

PetID - Unique hash ID of pet profile
AdoptionSpeed - Categorical speed of adoption. Lower is faster. This is the value to predict.
Type - Type of animal (1 = Dog, 2 = Cat)
Name - Name of pet (Empty if not named)
Age - Age of pet when listed, in months
Breed1 - Primary breed of pet (Refer to BreedLabels dictionary)
Breed2 - Secondary breed of pet, if pet is of mixed breed (Refer to BreedLabels dictionary)
Gender - Gender of pet (1 = Male, 2 = Female, 3 = Mixed, if profile represents group of pets)
Color1 - Color 1 of pet (Refer to ColorLabels dictionary)
Color2 - Color 2 of pet (Refer to ColorLabels dictionary)
Color3 - Color 3 of pet (Refer to ColorLabels dictionary)
MaturitySize - Size at maturity (1 = Small, 2 = Medium, 3 = Large, 4 = Extra Large, 0 = Not Specified)
FurLength - Fur length (1 = Short, 2 = Medium, 3 = Long, 0 = Not Specified)
Vaccinated - Pet has been vaccinated (1 = Yes, 2 = No, 3 = Not Sure)
Dewormed - Pet has been dewormed (1 = Yes, 2 = No, 3 = Not Sure)
Sterilized - Pet has been spayed / neutered (1 = Yes, 2 = No, 3 = Not Sure)
Health - Health Condition (1 = Healthy, 2 = Minor Injury, 3 = Serious Injury, 0 = Not Specified)
Quantity - Number of pets represented in profile
Fee - Adoption fee (0 = Free)
State - State location in Malaysia (Refer to StateLabels dictionary)
RescuerID - Unique hash ID of rescuer
VideoAmt - Total uploaded videos for this pet
PhotoAmt - Total uploaded photos for this pet
Description - Profile write-up for this pet. The primary language used is English, with some in Malay or Chinese.

Unstructured Data

Image Related

Raw Images
Metadata: Annotation, Dominant Colors, Crop Hints

Text Related

Sentiment: Sentences, Documents
Entities
Language

Data Exploration

Categorical Features: [ 'Type', 'Name', 'Breed1', 'Breed2', 'Gender', 'Color1', 'Color2', 'Color3', 'Vaccinated', 'Dewormed', 'Sterilized', 'Health', 'State', 'RescuerID', 'PetID' ]
Continuous Features: [ 'Age', 'MaturitySize', 'FurLength', 'Quantity', 'Fee', 'VideoAmt', 'PhotoAmt', ]
Textual Features: [ 'Description' ]
Image Features: []

Target Feature: 'AdoptionSpeed'

Histograms

Preprocessing

Normalization

Power law distribution : [ 'Age', 'FurLength', 'Quantity', 'Fee', 'VideoAmt', 'PhotoAmt' ]
Gaussian Distribution = ['MaturitySize']

Categorical Embedding

Categorical Features:

'Type': 2 -> 1
'Name': 9060 -> 100
'Breed1': 176 -> 50
'Breed2': 135 -> 50
'Gender': 3 -> 2
'Color1': 7 -> 3
'Color2': 7 -> 3
'Color3': 6 -> 3
'Vaccinated': 3 -> 2
'Dewormed': 3 -> 2
'Sterilized': 3 -> 2
'Health': 3 -> 2
'State': 14 -> 5
'RescuerID': 5595 -> 100
'PetID': 14993 -> 100

Embedding Architecture: categorical features -> embedding -> concat -> fc(100) -> fc(100)

Visualization with t-SNE

Textual Features

doc2vec, word2vec

Image Features

CNN feature encoding (ResNet)

Model

Gradient Boosting Tree (LightGBM)

Neural Network (PyTorch)

Multilayer Perceptron

Training

Stratified K-Fold

Used stratfied 5-fold to obtain indicies for training (11994/80%) and validation (2999/20%) sets.

To generate a smaller dataset, stratified 5-fold was apply to one of the validations sets (2999/20%) to obtain smaller training (2399/16%) and validation (600/4%) sets

Result

Metric

Quadratic Weighted Kappa: 0.259

References

Airbnb: GDBT -> Deep Learning
Entity Embeddding of Categorical Variables
ResNet
DenseNet
doc2vec