Скачать с ютуб Uplift Modeling to Detect Causal Effect | Uplift Modeling & Causal Inference to Improve Instagram в хорошем качестве

Uplift Modeling to Detect Causal Effect | Uplift Modeling & Causal Inference to Improve Instagram

#datascience #machinelearning #artificialintelligence #analytics #statistics Uplift modelling is still an active area of research in the data science community, but its practical applications are quickly gaining popularity in various domains. While a propensity model can just emit chances/probability of an event happening while an uplift model goes a step further to reveal the uplift in the event’s propensity that can come due to the applied changes. Some of the common use cases where uplift modeling shines compared to a propensity model are the following 1. A New Health Treatment: Uplift modeling can help understand how treatments might impact certain groups differently. 2. Cross-Sell: A company may want to run a cross-sell campaign with a limited budget and target the most likely-to-buy customers. 3. User Experience Testing: A company may want to understand how the change in the website or the app affects different types of people differently. In order to create an Uplift model, we must have data from a randomized control trial in which a random set of customers were targeted with the intervention (the treatment group) and another random set of customers were not targeted (the control group). An Uplift model uses this information to differentiate between four types of users (persuadable, sure-things, lost causes & sleeping dogs). Some techniques to create Uplift Models are. 1. Direct Uplift Models a. Class Transformation b. Regression Transformation 2. Indirect Uplift Models or Meta-Learners a. S-Learner b. T-Learner c. X-learner Instagram uses Uplift Modeling to efficiently manage its notifications. It improves user experience by sending limited notifications to only customers which are more likely to get active on the app avoiding sending unwanted notifications to users who are organically active in the app and anyways will see the content. In this video, we also cover how is the model trained to detect uplit and how in real-time it decides which customers to send or drop the notifications to. #datascience #analytics #machinelearning Blog Link: https://engineering.fb.com/2022/10/31... About DataTrek Series    • Introduction to DataTrek: Data Scienc...   Business Enquiries: [email protected] Find me on Instagram:   / simplyspartanx   Music: https://www.bensound.com/royalty-free...