Please enable JavaScript.
Coggle requires JavaScript to display documents.
1. Input data, 2. EDA, 3. ML model, 4. Different Type of Output - Coggle…
1. Input data
conclusive data(for off-line)
duration
whole page
total duration of a page
each region
duration of fixation in Q
duration of fixation in C
duration of fixation in P
fixation
mean
whole page
count of fixation in whole page/time spent in whole page
each region
count of fixation in P/time spent in P
count of fixation in Q/time spent in Q
count of fixation in C/time spent in C
count
each region
count of fixation in Q
count of fixation in C
count of fixation in P
whole page
count of fixation in whole page
count of fixation/count of eye
procedural data(for real-time)
coordinates
ConLSTM
sequence
raw data
類似股票預測time series的感覺吧？？？
[[x1,y1], [x2,y2], [x3,y3], [x4,y4], [x5,y5]....]→ predict label
region
without duration
[Q, Q, P, P, C]
with duration
[Q, CCC, P, PP]
Topic branch
2. EDA
region
count compare
duration compare
per question analysis
maybe delete some unsuitable question
per person analysis
maybe delete some unsuitable person
distribution
:check: data balance
3. ML model
Superviesd learning
normal feature
sequence feature
Unsupervised learning
classify questions
picture and text related
focus on picture and text relaation
picture and text not related
maybe only use text for future prediction
classify people
4. Different Type of Output