C161 Proj
Isaiah C. Mireles
2025-10-21
User Interest Model
Objective :
- Enhance ads click-through rate (CTR) prediction accuracy by leveraging ad logs, user profiles, and cross-domain data.
Target Domain :
- Because People are less likely to click on ads. When you open an ad you either click the link or just click off. Sparse data.
Source Domain :
More Data about peoples behaviors online. When on apps you like spending time on an app. More enriched data.
clean data & data analysis : Feature Engineering. Do feature selection.
CTR : \(\frac{\text{Total Clicks}}{\text{Total Impressions}}\)
Objective : Choose Relevant Features
Source : How many Unique values does each feature have?
## ct
## e_section 2
## label 2
## cillabel 2
## i_dtype 5
## u_phonePrice 7
## u_browserLifeCycle 8
## u_browserMode 8
## u_feedLifeCycle 8
## u_refreshTimes 10
## i_dislikeTimes 10
## i_upTimes 10
## e_ch 19
## e_po 27
## pro 85
## e_rn 99
## i_cat 208
## e_m 274
## i_regionEntity 378
## u_newsCatDislike 810
## e_pl 3113
## e_et 3663
## i_s_sourceId 4040
## i_entities 92747
## i_docId 98600
## u_userId 180123
## u_newsCatInterests 202636
## u_newsCatInterestsST 625044
## u_click_ca2_news 758921Source Domain : Categorical & Numerical Variables :
Categorical :
e_section : Information flow scenario type — indicates the section or context within the app where the user’s interaction with news content occurred (for example, main feed, hot news, video section, topic channel).
label : Indicates whether a user clicked on the ad (or content) — it is the target variable for CTR prediction (For ex. 1 = clicked, 0 = not clicked)
## $e_section
##
## $label
##
## $cillabel
##
## $i_dtype
##
## $u_phonePrice
##
## $u_browserLifeCycle
##
## $u_browserMode
##
## $u_feedLifeCycle
##
## $u_refreshTimes
##
## $i_dislikeTimes
##
## $i_upTimes
Target : How many Unique values does each feature have?
## ct
## site_id 1
## label 2
## gender 3
## app_score 3
## city_rank 4
## inter_type_cd 4
## net_type 6
## series_group 7
## age 8
## u_feedLifeCycle 8
## creat_type_cd 9
## u_refreshTimes 10
## app_second_class 20
## series_dev 27
## emui_dev 27
## residence 35
## hispace_app_tags 43
## slot_id 60
## spread_app_id 116
## device_name 256
## city 341
## adv_prim_id 545
## device_size 1547
## ad_close_list_v003 1715
## ad_close_list_v002 2701
## ad_close_list_v001 3883
## pt_d 5436
## task_id 11209
## adv_id 12615
## ad_click_list_v003 61102
## user_id 65297
## ad_click_list_v002 95376
## ad_click_list_v001 108720
## u_newsCatInterestsST 187576
## log_id 1176633