Picture

– The product driven by science consists of several aspects that every leader needs to be considered

– The machine learning algorithm is one part of the whole – we need to consider things like interpretability, trade-off between calculation and accuracy costs, among others

Introduction

Practical data science training is a multi-dimensional field. The machine learning algorithm is basically one part of the entire project driven by data-to-end. I often find fans of initial data science that don’t have a complete picture in mind during their early days.

There are several practical considerations that we need to be accountable while building solutions driven by science for real-world situations. And here’s a strange part – it’s not only limited to the data side of the thing!

Some more important warehouses in projects driven by scientific science include:

– Seamless user experience

– Exchange between computing costs and accuracy of our machine learning models

– How well we interpret the model, among others

 Table of contents

– Fast recap of what we have discussed in this data science leader series

– Data science is part of the whole

– Synergy with the user interface module (UI) and overall user experience (UX)

– Trade-off between costs and computational system accuracy

– Interpretability model

Fast recap of what we have discussed in this data science

In the first article, we talk about the three main constituents whose purpose must be harmonized for the success of the development and dissemination of products driven by data. This constituency is:

– Teams that face customers

– Executive team

– Data Science Team

This is basically the main stakeholder in the science project. You can read the full article, and detailed damage, from each stakeholder here.

In the second article, we discuss ways to bridge the gap between qualitative business requirements and quantitative input for machine learning models. In particular, we talk about defining the criteria for the success of the product driven by data in a way that progress can be measured in real quantitative terms.

This article also provides a framework for capturing the right data granularity with a consistent human label to train accurate machine learning models. Finally, the article reflects how the right composition of the team is very important for final success to the end.

Data science is part of the whole

Take a moment to think about some products driven by the data you use regularly. There is a good chance you might think of one of the following:

– Online search engines that provide relevant responses to your search terms but also help you improve your search when you enter a query

– Word processors that check spelling and grammar construction from your text and are fine or make recommendations for correction

– Social media platform that personalize the content or person you must connect based on relevance with your interaction on the platform

– E-commerce portal which recommends what you have to buy based on your current shopping basket and / or your shopping history

Two special examples are:

Financial loan institutions have developed solutions driven by data that decides whether the applicant deserves a loan. If yes, what is the optimal loan amount?

The very large data center with thousands of computer crunching numbers servers for various critical business requirements has developed solutions that are driven by data that analyze the log crossing server, database, and network traffic to predict which server will be turned on (or turns off) and how to manage the cooling unit.

Let’s understand this using the examples we mentioned above:

Say search engines make real-time recommendations to enhance search requests because users type. IR components are worthless if they cannot restore significant information in a few seconds

For data center management solutions, the IR component does not need to bother about speed but must weigh its output based on its impact on the continuity of the client’s business that it is filed

In the loan solution mentioned above, the IR component must pay special attention not to accelerate the execution or ‘business wisdom’ but the most important thing to ‘explained and fair in the eyes of the authorities regulating’

Users interact with the final application through the user interface (UI). User experience (UX) must be designed in a synergistic way with the strength of the scientific component of the underlying data while disguising its shortcomings.