Datascience vs Machine learning
November 16, 2017
Reading
Add Comment
Datascience vs Machine learning
Data science is all about making use of existing data to better understand a behavior, a phenomenon or to predict future insights.
As you can see in the picture, data science requires a variant set of skills that collectively contribute to the ultimate goal: making use of data. Machine learning is the part that requires math/statistics knowledge and the skills to code those algorithms to build models of the existing data.
Machine learning contributes to data science largely in all prediction and classification problems.
DS is more like a practice and given that it is a relatively new buzzword for describing the task of deriving insights from data regardless of volume, variety, velocity and space towards actionable decision-making, it is hardly used in the theoretical contexts. So until recently you would hardly come across academic or research programmes in DS. This is not the case for ML. While DS encompasses all phases of end-to-end data analytics (e.g. preprocessing, analysis, validation, interpretation, and deployment) ML on the other hand lay more emphasis on the techniques used in the analysis, validation and interpretation phases.
In academic research, ML researchers, mostly with backgrounds in computer science, mathematics, statistics or physiscs, develop new clustering, regression, model selection and validation algorithms for example, or improve on existing techniques such as devising novel strategies for determining optimal deep network structures. DS researchers on the other hand may have diverse technical backgrounds and are often concerned with applying statistical and ML techniques to address problems in diverse domains.
In the industry, the domains of DS and ML is delicately overlapping with rather fuzzy boundaries. Early adopters will have job roles such as Machine Learning Engineer or Scientist to refer to a whole lot of responsibilities requiring varying technical competence. Recently companies we have seen separation of concerns or responsibilities leading to more specific roles e.g. Data Engineer--engineering systems and tools to move/store data and deployment of predictive models from the data scientists, Data Scientist-- identfying organizational data needs and gaps, building predictive models based on highlevel organization goals and performing analytics on product data using ML techniques, carry out experiments e.g. A/B testing, converting insights to actionable recommendation and reports for high-level managers.
Below, I link a figure depecting the intricacies of DS, ML and other allied fields and how it fits into problem solving in many organizations.
0 comments:
Post a Comment