Paper: arXiv 2306.12969

Authors: David Noel

Abstract

This paper will analyze and implement a time series dynamic neural network to predict daily closing stock prices. Neural networks possess unsurpassed abilities in identifying underlying patterns in chaotic, non-linear, and seemingly random data, thus providing a mechanism to predict stock price movements much more precisely than many current techniques. Contemporary methods for stock analysis, including fundamental, technical, and regression techniques, are conversed and paralleled with the performance of neural networks. Also, the Efficient Market Hypothesis (EMH) is presented and contrasted with Chaos theory using neural networks. This paper will refute the EMH and support Chaos theory. Finally, recommendations for using neural networks in stock price prediction will be presented.

Complexity vs Empirical Score

  • Math Complexity: 6.0/10
  • Empirical Rigor: 5.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper employs advanced mathematical derivations for the Levenberg-Marquardt algorithm and NARX models, placing it in the high math category. It also outlines a specific implementation plan with data acquisition, training parameters, and performance metrics, giving it moderate empirical rigor.

Research Flowchart

  flowchart TD
  A["Research Goal<br>Refute EMH & Support Chaos Theory"] --> B{"Data & Inputs"}
  B --> B1["Historical Stock Data<br>Time Series"]
  B --> B2["Contemporary Methods<br>Fundamental/Technical/Regression"]
  
  B1 & B2 --> C["Methodology<br>Dynamic Neural Network DNN"]
  
  C --> D["Computational Process<br>Pattern Recognition in Non-Linear Data"]
  D --> D1["Training DNN on Time Series"]
  D --> D2["Comparing DNN vs Contemporary Methods"]
  
  D1 & D2 --> E{"Key Findings & Outcomes"}
  E --> E1["DNN Outperforms Current Techniques"]
  E --> E2["Refutation of Efficient Market Hypothesis EMH"]
  E --> E3["Validation of Chaos Theory"]
  E --> E4["Recommendations for DNN Use in Prediction"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes implementing a time series dynamic neural network to predict daily closing stock prices.

Source sentence
This paper will analyze and implement a time series dynamic neural network to predict daily closing stock prices.

Assumes 3

  • The network uses 70% of the data for training, 15% for validation, and 15% for testing. data
    Source sentence
    Furthermore, 70% of the input dataset is used for training, 15% for validation, and 15% for testing.
  • The closing price is both a feedback signal and an input/output target. model
    Source sentence
    which is the closing price of the stock, is a feedback signal and is also an input and an output (target).
  • Predictions are evaluated without accounting for trading frictions or slippage. execution
    Source sentence
    The network errors decreased until iteration 5 (out of 11 iterations), where their gradients were no longer changing.

Data

  • Yahoo finance stock data · finance.yahoo.com market-prices public
    01/01/2010 – 03/31/2014
    Source sentence
    Four years (01/01/2010 – 03/31/2014) of prior stock data would be retrieved from finance.yahoo.com using a custom‑written function based on MATLAB data‑feed plugin.

Costs

Not applicable The paper reports no trading or execution result.

Finds 3

  • The model achieves a correlation of approximately 0.998 between outputs and targets, indicating very accurate predictions. predictive-accuracy
    Source sentence
    A number equal to 1 signifies a perfect correlation between targets and outputs. In our model, the value is approx. 0.998, which signifies a very accurate correlation.
  • Maximum prediction error over 100 simulated days is 1.17%, with errors fluctuating around zero. performance
    Source sentence
    The maximum error difference recorded was 1.17%, indicating that the predicted (simulated) and target (actual) values are sufficiently close to each other to be deemed negligible.
  • Error autocorrelation shows only the lag‑0 value is significant, confirming that most errors are white noise. statistical
    Source sentence
    For our model, all the bars except the one at Lag 0 fall within the 95% confidence limit of around 0, which validates this model as very accurate.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 11 related papers). Coverage is partial, so this is neither replication nor refutation.

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.