Learning Reflections 2019

The recent Singapore Fintech Festival in November 2019 covered number of aspects of Data Science and Analytics as part of the talks and product exhibits. 

The Deputy Prime Minister of Singapore referred to the quote; “Technology is never good or bad neither is it neutral” – Melvin Kranzberg, in the context of describing 3 things that need to be considered in the use of technology: People focus, staying open & connected and good governance of new technologies. This made me reflect on the last few months of learning on Data Science & Analytics at the Berkeley Haas Executive Education course.

Motivation

✥ Today’s business environment that involves substantial use of technology & application of econometric theory as part of different aspects of Data Science & Analytics, one needs to have a strong comprehension of what it involves in order to make informed decisions. The ask is not to be a data scientist but to have sufficient depth of knowledge and understanding of the topic that is disruptively changing the way the societies function and businesses operate. One needs to better understood the topic to make positive decisions and avoid the pitfalls due to ignorance.

✥ The goal was to learn and build capabilities so as to:

• judge what good data analytics looks like (Understand the What & How things are done);
• identify opportunities where analytics adds value (Where & When these concepts can be applied); and
• lead with confidence and know what works best with data science team structures.   

✥ Specifically, the objective was to learn how to frame a business problem into a data science problem, recognise what a good data scientist should be like and engage with Data Science teams to solve business issues and problems.

✥ A personal goal was also to keep an open mind and avoid the trap of the experience bubble- extrapolating the past in the future.

Method & Mechanics 

✥ To achieve the objectives, a 6-month program with a mix of virtual and in person lectures seemed to be ideal. It seemed to have a a right balance between the shallow end (Novice) and deep end (Data Scientist) of the topic.

✥ The program with Berkeley-Haas covered:

• a good foundational learning about Economics-Statistics and Coding.
• site visits to a mix of established & startup companies in the data science value chain & companies from different industries, to reinforce the classroom learning and learn about practical application of theory.
• learning and application of python on Jupyter notebooks to get hands-on experience of tools required for data cleansing, transformation & visualisation, statistical modelling, numerical simulation & regression analysis.

✥ We learnt a number of topics about Statistics and Economics (Regression models-diagnostics, Behavioural economics, Causation & Prediction, Logistic models, Auction strategies and terms such as confusion matrix, area under the curve etc.).

Reflections-Message

1.    While Data Science involves use of technology-algorithms; the ability to contextualise-problems and solutions to arrive at good decisions is key for success.

📌 “Human Inputs” and experience are important to provide context, guidance, vet the validity of the problem and solution options.

📌 While Data Science involves use of technology-algorithms; the ability to contextualise-problems and solutions to arrive at good decisions is key for success.

📌 Measurement without management judgement and context can be dangerous and create an illusion of insight.

📌 While one can use data science and analytics to test hypothesis and answer known questions, the real value is in using data to raise unraised- hidden questions. For example; when data analytics was used to train a model for a defined question, it accidentally highlighted discrimination and bias in hiring, thus raising the question on the topic of bias and discrimination in the past. 

2.    Data is key but data without quality & context results can be misleading

📌 Throwing the “kitchen sink”- lots of data to a problem does not necessarily mean a more effective solution. One needs to support this with a view of the “model of the world”.

📌 Large size of data creates the challenge of spurious correlations due to large explanatory variables. One needs awareness to distinguish between correlation and causation while making conclusions and evaluating analysis.

📌 While Data does support objective decisions and solutions, it can also be used to define problem itself in a more objective manner.

📌 Good decision through data driven actions has certainty. However good outcomes by itself are uncertain.

📌 Apply Data Science concepts and tools to evaluate and question norms - critical thinking. Don’t take opinions as facts - test it with data and experiment to get data.

📌 When data is used to find patterns and make predictions, it may sometime provide counter-intuitive results that may reveal human biases & fallacies that consciously or unconsciously creep into human decisions. 

📌 Machine learning is not “magic”. Machine learning best describe the relationship between inputs and outputs for a collection of existing data. However; quality and change in inputs or a bit of pattern can have significant impact on the output – prediction.

3.    If risk management is about decisions – knowledge of data science and analytics is important; to understand the basis of management decisions; and know how data science and analytics can itself be used for better risk management.

The program not only helped provide the required confidence in dealing with Data Science vendors-projects and teams but also helped approach management challenges in a more objective-rationale manner.

While the technical aspects of the learning were great and helped in identifying problems and decision making, the richness of learning experience was far greater due to collaboration between people from diverse backgrounds (cultures, geographies, professional experience and education). This helped in perceiving things through different lens and arriving at newer solutions.

Thank You: Julien Oet, Happy Mor Lomigo, Kaushal Shah, Sukruth Pattanagiri, Audrey Shum, Chamika Godmanna, Matthias Zutter, Patrick Storey, Prabhat Shrivastav, Datuk Khan Mohd Akram Khan, Tuyen Kin, Ameer Dairi, Jenina Tanada, Maria Canziani, Marcello Di Maulo, Shabnam Azizian, Sharulnizam Sarip, Sultan Aldainy, Trinh Lee, Anastasia (An) Dy, Pradnyaditya Hendradi, Maha Muraish, and of course our professors/guides Sachar Kariv, Steve Tadelis, Keely Takimoto and Marose Eddy for the experience and the knowledge shared.