AIVestor Integrates Deep Learning Architecture to Build Global Market Trend Recognition Model
In an era where artificial intelligence continues to reshape the landscape of the financial industry, ETERNAL DIGITAL FUND LTD announced in July 2021 that its self-developed intelligent investment decision system, AIVestor, has officially integrated a deep learning architecture and successfully built a new-generation global market trend recognition model. This technological breakthrough marks a new stage in ETERNAL’s intelligent investment framework and introduces a higher-dimensional approach to analysis and decision-making for the global asset management industry.

AIVestor is an AI-driven investment decision system developed by ETERNAL DIGITAL FUND since 2019, designed to achieve trend detection and risk recognition across multi-asset markets through algorithmic self-learning and data-structure evolution. The introduction of deep learning represents another major upgrade following the establishment of the system’s quantitative framework in 2019, the deployment of the AI Risk Matrix in 2020, and the completion of prototype testing in early 2021. By adopting a hybrid model combining Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks, AIVestor has gained the ability to simultaneously interpret both the “current behavior” and the “historical inertia” of markets — enabling it to analyze complex cross-cycle financial signals with higher precision.
Bryan Thomas Whalen, Founder and Chief Investment Officer of ETERNAL DIGITAL FUND, stated:
“Market trends are no longer mere extensions of price movements — they are dynamic equilibriums formed by thousands of micro variables. Our goal is for AIVestor to understand the logic behind the markets, not just identify fluctuations in price.”
He further emphasized that the integration of deep learning allows the system to extract structural signals from heterogeneous global datasets, uncovering internal market cycles and emotional inflection points.
The new-generation AIVestor model was trained on market data from 28 major economies, covering a wide range of asset classes including equities, bonds, commodities, foreign exchange, and digital assets. Every 24 hours, the system extracts feature variables from more than 120 million structured and unstructured data points, performing signal attribution and trend reinforcement learning through a multi-layer neural network. Notably, for the first time, the system incorporates financial news and policy semantics into its training core. Using Natural Language Processing (NLP) models, AIVestor quantifies policy tone, market consensus, and investor sentiment — enabling true emotion-driven trend recognition.
According to preliminary testing by ETERNAL’s research laboratory, the new model achieved an average 9.6% improvement in trend prediction accuracy across multiple markets compared with the previous version. Additionally, its response speed in identifying extreme market turning points — such as sudden policy shifts or geopolitical events — improved by nearly 40%. Over the past six months of backtesting, AIVestor successfully captured several key signals, including the sectoral reversal of U.S. growth stocks, short-term yield adjustments in European bonds, and capital flow shifts in Asian markets.
The strategic significance of this system upgrade extends far beyond the realm of technology. As the global economy transitions into a phase of recovery and rebalancing following the pandemic, market volatility and structural transformation are becoming the new normal. The AIVestor Global Trend Recognition Model empowers ETERNAL to identify structural opportunities earlier in cross-market allocation — such as capital reflow into emerging markets, sector rotations between energy and technology, and the long-term influence of inflation expectations on bond markets.
