STACK-X Data Science Connect

This presentation gives an overview of MOM’s data strategy and data culture that support its transformation efforts

Analytics-readiness is something that all projects should have on Day 1 of its launch. The key to that is to avoid analytics debt. The data collected from policy implementations, or programme executions should address questions that will help finetune refinements down the road. The data generated from products and platforms should help inform how well it is working, and what new features to build, or what to tweak. But this is often not the reality. Many projects face challenges with analytics downstream, potentially leading to disappointments from both from the business side, as well as the data science practitioners. In this sharing, the speaker will share more about Analytics By Design, which is a methodology that GovTech had mooted, and is piloting internally.

This presentation gives an overview of how DSAID's Quantitative Strategy team has been actively supporting WOG data transformation by proliferating data-driven decision making in agencies. We will share about some of the high impact projects that the team has delivered over the years, covering a wide range of business domains, use cases and techniques. We will also discuss our philosophy, approach and methodology of using data and analytics and how we customise our service offering based on agency needs and data maturity to drive sustainable impact.

A multilevel model is a regression on grouped data in which parameters (i.e., the intercept or slopes/regression coefficients) are randomly varying and given a probability model. We will learn how it is differs from classical regression and briefly explore how multilevel modelling can be used in the context of schools.

In this presentation, we review recent trends & discussions concerning concepts in Statistical Modelling and Machine Learning. We will also delve into the difference between inference & prediction and discuss strengths and concerns with some of the popular ideas that has evolved in the last century (e.g., use of p-value).

We have started the journey to transform GovTech into a model data analytics-ready organisation that is exemplary in the use of data! This presentation will give an introduction on GovTech’s Data Strategy and the various enablers that have been put in place to empower GovTechies to make use of data to improve their operations and gain insights for policy making.

This brief discusses how we have enabled user-centric data sharing, concierge services and self-help exploitation of sensor domain core data in our of the four Trusted Centres under the Government Data Architecture: 1) Foundational governance underpinning our approach to data request flow; 2) Importance of leveraging on an enterprise identity provider store (IdP) and its influence on our data fabric; 3) Enabling user-led discovery and exploitation with Databricks.

Data and AI have become strategic priorities for every organisation, but today's complex and outdated legacy infrastructure struggles to deliver on the promise of AI. It's time for a new data architecture build to meet your needs today - and future-proofed so it's ready for whatever tomorrow brings. Find out how the Databricks Lakehouse combines the best of both data warehouses and data lakes with a lake-first approach.

Data engineering strategy cultivates organizations to plan out their successful data journey. DSAID’s Data Engineering team has been single-minded in our mission to develop central platforms to support data sharing, analytics and data engineering capabilities, as well as partnering agencies to co-create critical data infrastructure and uplift their in-house data engineering capabilities. In this talk, we will cover components of the data engineering strategy which includes knowing the right processes & data, relevant tools & techniques, latest technology & infrastructure as well as some of our exciting work-in-progress.

National Economic Research & Visualisation Engine (NERVE) serves as the Sectoral Data Hub for the economic domain, facilitating access to identifiable firm-level, industry-level and macro-level data across government agencies. It draws data from both within the government as well as the private sector to enhance economic surveillance, inform strategy formulation, evaluate policy outcomes and facilitate service delivery.

This presentation gives an overview of the Artificial Intelligence (AI) projects Govtech has done for various agencies as well as the approach taken in supporting government AI use cases. In Singapore, AI is applied in a variety of uses cases, ranging from policy to ops work. Examples include text analytics, speech to text, fraud detection, and video analytics. Essentially GovTech aims to excite officers on the possibilities of AI and get them to think of how they can apply AI to the domain problems they have at hand.

The gap between the promise of AI and the value it brings to government agencies is closing but still wide. In this presentation, Chandler McCann will share the technical and practical approaches to model development and production that have led to massive results across the U.S Government with automated machine learning.

Eugene will discuss various applications of AI in cybersecurity, from detecting machine-generated phishing emails to automating code review. He will highlight established use cases as well as cutting-edge research.He will also examine the latest proposed standards in the use of AI on cybersecurity.

The benefits of AI are widely known today. Increased interest among business users to adopt AI-enabled solutions is pushing the technology’s frontier on a daily basis. The rate of innovation and translation in this space is fast, and advancements in AI are being rapidly productized for commercial applications. An example is AI in the field of Video Analytics. Learn how the VA team at GovTech collaborate with industry to keep pace with advancements in AI and develop solutions for public sector agencies.

The Video Analytics (VA) team builds deep learning models to address specific public sector use cases. As Deep Learning models become more advanced with increasing complexity, the need for model interpretability has come into greater focus. Understanding features that contribute to a Deep Learning model's output helps AI Engineers design better models and troubleshooting. There are methods known today to help with Deep Learning model interpretability, such as those requiring visual explanations on local prediction instances. But we aspire to create a global metric to quantify explainability. Find out how we are doing this in our collaboration with Meta/PyTorch.