Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
41 tokens/sec
GPT-4o
59 tokens/sec
Gemini 2.5 Pro Pro
41 tokens/sec
o3 Pro
7 tokens/sec
GPT-4.1 Pro
50 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Listening to Chaotic Whispers: A Deep Learning Framework for News-oriented Stock Trend Prediction (1712.02136v3)

Published 6 Dec 2017 in cs.SI, cs.LG, and q-fin.ST
Listening to Chaotic Whispers: A Deep Learning Framework for News-oriented Stock Trend Prediction

Abstract: Stock trend prediction plays a critical role in seeking maximized profit from stock investment. However, precise trend prediction is very difficult since the highly volatile and non-stationary nature of stock market. Exploding information on Internet together with advancing development of natural language processing and text mining techniques have enable investors to unveil market trends and volatility from online content. Unfortunately, the quality, trustworthiness and comprehensiveness of online content related to stock market varies drastically, and a large portion consists of the low-quality news, comments, or even rumors. To address this challenge, we imitate the learning process of human beings facing such chaotic online news, driven by three principles: sequential content dependency, diverse influence, and effective and efficient learning. In this paper, to capture the first two principles, we designed a Hybrid Attention Networks to predict the stock trend based on the sequence of recent related news. Moreover, we apply the self-paced learning mechanism to imitate the third principle. Extensive experiments on real-world stock market data demonstrate the effectiveness of our approach.

A Deep Learning Framework for News-oriented Stock Trend Prediction

The research paper "Listening to Chaotic Whispers: A Deep Learning Framework for News-oriented Stock Trend Prediction" explores the challenges and opportunities in predicting stock market trends based on online news content. This proposition hinges on the application of advanced deep learning methodologies to overcome the highly volatile and non-stationary attributes of financial markets. As traditional stock trend prediction methods often falter in capturing dynamic market shifts, this paper taps into burgeoning NLP and text mining capabilities to analyze news-based influences on stock prices.

The paper introduces a Hybrid Attention Network (HAN) designed to model human-like learning by incorporating principles of sequential context dependency and diverse influence from the news. These principles reflect human investors' tendencies to synthesize sequences of relevant news and assess their varying impacts rather than relying on isolated news articles. Central to this framework is the hybrid attention mechanism, which includes both a news-level attention mechanism that discerns the relative importance of individual news items and a temporal attention mechanism that evaluates the significance of news across different time points.

The implemented HAN exploits these dual-attention mechanisms to produce a composite understanding of stock trends. By leveraging the self-paced learning (SPL) mechanism, the framework further refines its prediction capabilities, progressively learning from challenging samples over the training process. This SPL integration enables adaptability in learning sequences, providing a means to effectively incorporate and respond to variable market conditions.

Extensive empirical tests on Chinese market data from 2014 to 2017 underline the effectiveness of this approach. The HAN framework, especially when augmented with SPL, demonstrates superior performance in trend prediction accuracy compared to more conventional methods such as Random Forests and Multi-layer Perceptrons. In measuring real-world applicability, the research simulates a trading strategy utilizing the framework's predictions. The simulation results exhibit enhanced annualized returns, suggesting potentially significant implications for stock market investors seeking to leverage information from news media.

These findings contribute valuable insights into the synthesis of NLP techniques and stock market analysis. By successfully navigating the noise of chaotic online content through a structured, neural network-based approach, the paper underscores a potential avenue for the operational deployment of AI-driven predictive tools in financial markets.

Looking forward, the implications of this research suggest further exploration into multi-faceted news analytics through industrial interconnections and enhanced integration with traditional technical analysis methods. Such augmentations could potentiate an even more comprehensive predictive apparatus adaptable to disparate and multifarious data sources.

User Edit Pencil Streamline Icon: https://streamlinehq.com
Authors (5)
  1. Ziniu Hu (51 papers)
  2. Weiqing Liu (36 papers)
  3. Jiang Bian (229 papers)
  4. Xuanzhe Liu (59 papers)
  5. Tie-Yan Liu (242 papers)
Citations (286)