Welcome and start of conference
After the excitement of alternative data, new AI and ML techniques and cloud elastic computing, what's actually been achieved in using these advancements in the financial services industry? A senior industry technologist shares his thoughts on the current state of play for data science and data engineering, what's been effective and what is over-hyped, and how to attract the right talent to solve these problems.
· Improving models to decrease the chances of false predictions
· Automation technology being utilised as part of crisis solution
· Opportunities opening up for investment strategies post-Covid
· While many firms experience massive losses as the market crashed earlier this year, a unique approach to AI allowed Castle Ridge Asset Management to continually deliver steady returns for their clients
Uncovering future sources of risk requires an understanding of how the world is connected through data. Leveraging data mosaics allows investors to glean new insights about real-time business risk. The future of alternative data is not found off-the-shelf. Curated creation of new insights is becoming the norm.
Birds of a feather session: Sentiment Scoring and Event Detection Using Neural Networks
There are numerous variations in deep learning-based methodologies used to perform sentiment analysis and event detection. In this workshop, we’ll demonstrate a relatively straightforward approach that is still quite effective. After briefly reviewing the deep learning basics, we’ll create a 4-layer neural network based on Long Short-Term Memory (LSTM) networks using the TensorFlow library and its Keras interface. With the help of publicly available pre-trained word embeddings, we will fine tune the model and train it to output a sentiment score and detected events.
Speaker: Marko Kangrga, Senior Data Scientist, RavenPack
Birds of a Feather Session: The Value and Certainty of Transaction Data During Uncertain Times
The COVID-19 pandemic turned the economy upside down and inside-out in 2929. Investment and business models built on years of regular, expected economic data have been thrown out the window and investors are grappling with ‘what’s the new normal?’ This session will highlight how transaction data from Facteus became a key ‘source of truth’ during the pandemic and beyond to enable investors to understand the consumer economy in real-time and make strategic investments. Examples of data sourcing, compliance, normalization, and industry/merchant analysis will be shared in this session.
Speaker: Don Wood, Data Strategist, Facteus
Birds of a Feather Session: Avoid spurious correlations and optimize your portfolio with Causal AI
The current state of the art in machine learning relies on past patterns and correlations to make predictions of the future. This approach can work in static environments and for closed problems with fixed rules. However it does not work for financial time-series and other dynamic systems. In order to make consistently accurate predictions about the future, and to achieve true artificial intelligence, the development of new science that enables machines to understand cause and effect is required. This talk will present the power of Causal AI and some practical applications that are already being used in the field.
Speaker: Darko Matovski, CEO and Co-Founder, CausalLens
· Building a model that can optimally hedge an option while reducing risk
· Examining the feasibility of using RL to hedge basis risk in a realistic setting
· How to personalize, maintain and monitor your models
· Using AI for complex and dynamic decision-making
· Building models to optimize and predict investment outcomes
· Bayesian interpretation of Q-learning
· Applications of Baysian Q-Learning to Finance
· Broader applications of Bayesian Deep Learning
· Mapping the recruitment landscape and how to ensure you remain competitive in hiring and retaining data professionals
· Where to find the best talent in the industry
· Outlining the infrastructural data strategy and a leading fund
· How to reliably build a data strategy from scratch
· Defining your data strategy and investing in the infrastructural elements to support it
· Building models to optimize and predict investment outcomes
- The future for the commoditization of data engineering
- Utilising cloud providers and open source tools to streamline your optimise your investment workflow
- Defining key aspects from whole investment workflow
· Identifying at what point human intervention is necessary
· Avoiding building models on spurious relationships
· Overcoming structural challenges and training models differently
· Driving value through cost reduction
· How to organize institutional data to scale
· Assessing the best cloud deployment models for your fund
· Challenges in switching to cloud computing
· Leveraging cloud services in a scalable, streamlined way
· Articulating a new strategy – testing new alternative datasets, new open-source techniques and machine learning processes
· Importance of organisations to allow for the development of creativity and innovation