IDS_2주차: Data Visualization

EDA, Explorative Data Analysis

: 다양한 차원과 값을 조합해가며 특이점이나 의미있는 사실을 도출하고 분석의 최종 목적을 달성해나가는 과정

 

1) Verify expected relationships

2) Find some Unexpected structure in the data

3) Deliver data-driven insights in the right questions and not bias the investigation

4) Provide the context around the problem

Main topic in EDA: 저항성의 강조 / 잔차계산 / 자료변수의 재표현 / 그래프를 통한 현시성 제공

 

 

 Graphs

Graphs are used for data exploration( visualization)

 Basic Plots

• Line graphs: line graphs are used for time series
• Bar charts: bar charts are used for categorical variables (범주형 변수)
• Scatterplots: scatterplots display the relationship between two numerical variables

 Distribution Plots

: Distribution plots display 'how many' of each value occurs in a data set 

  or for continuous data or data with many possible values, 

• Boxplots: 
  Top outliers defined as those above Q3 + 1.5(Q3 - Q1)
  max = maximum of non-outliers
  min = bottom outliers  
  IQR, Inter Quartile Range = Q3 - Q1
  IQR의 약 1.5배의 최소, 최대값에서 벗어나는 경우 이상치로 판별함
  
  Side-by-side boxplots are useful for comparing subgroups

• Histogram: Histogram shows the distribution of the outcome variable
• Heat Maps: Color conveys information.
  Heat Maps is used to visualize correlations and missing data in data mining.

 

Multidimenstional Visualizatoin 

• Scatterplot with color added
• Matrix Plot: Matrix shows scatterplots for variable pairs
• Rescaling to log scale
• Aggregation(집합)
• Scatterplot with labels