“Data science is a fusion of artificial intelligence and business knowledge.”
That’s the view of Hosein Alizadeh, data science team leader at Wesfarmers, who argues successful analytics programs require business people and data scientists to work collaboratively in order to realise the value data can bring to their business.
“The aim of data science is to extract actionable insights from data that address business problems,” Alizadeh said during a presentation at the Quest Ignite AI and Machine Learning Summit in Sydney yesterday.
That requires data scientists to understand business problems and on the flip side, business managers to improve their knowledge of AI and machine learning.
This year Wesfarmers established an advanced analytics centre to work with its divisions including Kmart, Officeworks and Bunnings, to build their out their advanced analytics capabilities.
Wesfarmers managing director Rob Scott has identified data and digital as a top three strategic priority for the diversified group and the goal is to grow the advanced analytics team from zero to between 15 and 20 data scientists, data engineers and business translators.
“We see a significant potential for our shareholders in leveraging our data assets,” Alizadeh said.
The core focus of the new department is enhance data-driven decision making among the divisions, Alizadeh said, which means being involved in everything from delivering use cases to implementation of machine learning models.
“First of all we need to understand and define the business problem. It is a very crucial and important part of the whole framework,” Alizadeh said.
Alizadeh argued if a data scientist is not engaged enough with the business in the early stages of an analytics project, they can still prepare the data and build a model but will struggle to deploy their work into the business.
“I sometimes hear from analytics managers or analytics teams that they’ve built a great model but the business doesn’t understand what it does and value of that isn’t appreciated,” he said.
“To me being a data scientist, I would say part of that might be because there is not enough business input into the data science framework or the whole project.”
The reverse is also true: business needs to increase their awareness and understanding of artificial intelligence.
During his presentation Alizadeh identified four types of machine learning algorithms that can be applied to business problems: descriptive algorithms (what happened), explanatory algorithms (why did it happened), predictive algorithms (what will happen) and prescriptive algorithms (what should we do to make X happen).
He argued business units need a high level understanding of four different types of algorithm and how they can be used be address business problems.
For example a prescriptive algorithm is used to recommend an action to achieve a certain outcome such as, what’s the optimal discount for a SKU [stock keeping unit] to sell out in next four weeks?