Submit Predictions
Post on 22-Feb-2016
39 Views
Preview:
DESCRIPTION
Transcript
Submit Predictions
Statistics &Analysis
Data Management
Hypotheses
Goal
Get Data
Predict whom survived the Titanic Disaster
+
Goal: Achieve High Prediction Score
Score = Number of Passengers in Test Dataset Correctly Predict Passenger’s Fate
Submit Predictions
Statistics &Analysis
Data Management
Hypotheses
Goal
Get Data
Predict whom survived the Titanic Disaster
Woman and Children First
Training and Test Data
Training Data
N=89139% Survived
Test Data
N=418All Titanic
PassengersN= 2,223
All Employees Subset of Current Employees
Subset of Current Employees
All Customers Subset of Customers
Subset of Customers
Develop Model
Variable Description Type Datapclass Passenger Class Categorical,
Ordinal1 = 1st; 2 = 2nd; 3 = 3rd
Pclass is a proxy for socio-economic status 1st ~ Upper; 2nd ~ Middle; 3rd ~ Lower
name Name TextSex Sex Categoricalage Age Numericsibsp Number of Siblings/Spouses Aboard Integer
parch Number of Parents/Children Aboard Integer
ticket Ticket Number Textfare Passenger Fare Numericcabin Cabin Textembarked Port of Embarkation Categorical C = Cherbourg; Q = Queenstown; S =
Southampton
Predictor Variables
Submit Predictions
Statistics &Analysis
DataManagement
Hypotheses
Goal
Get Data
Predict whom survived the Titanic Disaster
Woman and Children First
Read dataset into Excel, R, etc
Datasets: Training and Test
Develop Model Using Training Dataset and Apply to Test Data
Submit Predictions
Statistics &Analysis
Data Management
Hypotheses
Goal
Get Data
Predict whom survived the Titanic Disaster
Woman and Children First
Read dataset into Excel, R, etc
Some Age Missing Data, Analyze Gender Only
Gender Model
Training Data
Test Data
Develop Model
Submit Model
Leaderboard
320418
Submit Predictions
Statistics &Analysis
Data Management
Hypotheses
Goal
Get Data
Predict whom survived the Titanic Disaster
Woman and Children First
Read dataset into Excel, R, etc
Some Age Missing Data, Analyze Gender Only
74% Women, 19% Men
320 / 418 = 76.5%
top related