-
Notifications
You must be signed in to change notification settings - Fork 0
/
engineer-features-guide.txt
133 lines (75 loc) · 3.45 KB
/
engineer-features-guide.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
Understandable. Feature engineering can be challenging,
especially when you feel you've already included the most obvious factors.
Let's brainstorm some potential new features that could be relevant for predicting player
transfers in college football. I'll suggest ideas based on common reasons players might transfer
and factors that could influence their decision:
Playing Time Metrics:
Percentage of snaps played in previous season
Change in playing time from previous year to current year
Ratio of player's stats to team's total (e.g., player's passing yards / team's total passing yards)
Team Performance:
Team's win-loss record in previous season
Change in team's record from previous year
Team's conference ranking
Bowl game appearances or lack thereof
Depth Chart Position:
Player's position on depth chart (if available)
Number of players ahead on depth chart in same position
Coaching Changes:
Binary feature for head coach change in last year
Binary feature for position coach change
New coach's history of using transfer portal
Academic Factors (if available):
Player's major
Academic performance metrics
Availability of specific academic programs at current school
Geographic Factors:
Distance from hometown to current school
Number of schools closer to hometown in player's conference
NFL Draft Prospects:
Draft eligibility status
Pre-season draft rankings (if available)
Social Media Indicators:
Sentiment analysis of player's social media posts
Frequency of posts about team or school
Injury History:
Games missed due to injury in previous season
Cumulative injury metric over college career
Team Offensive/Defensive Scheme:
Compatibility of player's style with team's scheme
Changes in team's offensive/defensive coordinators
Competitive Metrics:
Quarterback rating relative to conference average
Performance in big games or rivalry matches
Financial Factors (for grad transfers):
Availability of scholarship for extra year
Potential NIL (Name, Image, Likeness) opportunities at current vs. potential schools
Team Discipline Issues:
Number of team suspensions or disciplinary actions
Player's personal conduct record
Redshirt Status:
Whether player was redshirted
Years of eligibility remaining
Transfer Portal Trends:
Number of transfers from player's position in previous year
Overall transfer rate for player's conference or division
Program Stability:
NCAA investigations or sanctions
Changes in athletic director or key administrative positions
Teammate Relationships:
Number of teammates from same high school or hometown
Turnover rate of players in same position group
Post-Season Performance:
Player's stats in bowl games or conference championships
Team's post-season success or failure
To implement these:
Assess data availability for each potential feature.
Create new columns in your dataset for the features you can reliably calculate or obtain.
Use domain knowledge to combine or transform existing features into more meaningful metrics.
Consider interaction terms between important features.
Normalize or standardize new numerical features as needed.
Use feature selection techniques (e.g., mutual information, feature importance from tree-based models)
to identify which new features are most predictive.
Remember, the goal is to capture factors that might influence a player's decision to transfer.
Even if you can't implement all of these, adding a few relevant new features could significantly
improve your model's predictive power.