[ad_1]
The authors share some important aspects of applied machine learning that are often overlooked in formal data science education.
![guillaume cory](https://miro.medium.com/v2/resize:fill:88:88/1*v9d5_08I0HrJaMEPX2E6Pg.jpeg)
![Towards data science](https://miro.medium.com/v2/resize:fill:48:48/1*CJe3891yB1A1mzMdqemkdg.jpeg)
YI know I’m leaning toward a clickbait title, but hear me out! I’ve managed several junior data scientists over the years and taught applied data science courses to master’s and PhD students for the past few years. . Although most of them have strong technical skills, we noticed some gaps when it comes to applying machine learning to real-world business problems.
Below are five factors I wish data scientists were more aware of in a business context.
- Think again about your target
- address imbalances
- Testing must be done in the real world
- Use meaningful performance metrics
- Is the score important?
I hope that reading this will help you advance your career as an intermediate-level data scientist.
This article focuses on a scenario where a data scientist is tasked with deploying a machine learning model to predict customer behavior. It is worth noting that this insight may also be applicable to scenarios involving product and sensor behavior.
LLet’s start with the most important thing of all.what‘ I’m trying to predict. Unless you focus on the right target, all subsequent steps (data cleaning, preprocessing, algorithms, feature engineering, hyperparameter optimization) will be in vain.
To be practical, targets should represent behavior rather than data points.
Ideally, the model fits the business use case and actions and decisions are taken based on its output. Ensuring that the targets used adequately represent customer behavior makes it easier for businesses to understand and leverage the output of these models.
Clothing retailer target example
[ad_2]
Source link