Comparative Study of Machine Learning Models for Stock Price Prediction
Authors:
Ogulcan E. Orsel,
Sasha S. Yamada
Abstract:
In this work, we apply machine learning techniques to historical stock prices to forecast future prices. To achieve this, we use recursive approaches that are appropriate for handling time series data. In particular, we apply a linear Kalman filter and different varieties of long short-term memory (LSTM) architectures to historical stock prices over a 10-year range (1/1/2011 - 1/1/2021). We quanti…
▽ More
In this work, we apply machine learning techniques to historical stock prices to forecast future prices. To achieve this, we use recursive approaches that are appropriate for handling time series data. In particular, we apply a linear Kalman filter and different varieties of long short-term memory (LSTM) architectures to historical stock prices over a 10-year range (1/1/2011 - 1/1/2021). We quantify the results of these models by computing the error of the predicted values versus the historical values of each stock. We find that of the algorithms we investigated, a simple linear Kalman filter can predict the next-day value of stocks with low-volatility (e.g., Microsoft) surprisingly well. However, in the case of high-volatility stocks (e.g., Tesla) the more complex LSTM algorithms significantly outperform the Kalman filter. Our results show that we can classify different types of stocks and then train an LSTM for each stock type. This method could be used to automate portfolio generation for a target return rate.
△ Less
Submitted 31 January, 2022;
originally announced February 2022.
Bound states at partial dislocation defects in multipole higher-order topological insulators
Authors:
Sasha S. Yamada,
Tianhe Li,
Mao Lin,
Christopher W. Peterson,
Taylor L. Hughes,
Gaurav Bahl
Abstract:
The bulk-boundary correspondence, which links a bulk topological property of a material to the existence of robust boundary states, is a hallmark of topological insulators. However, in crystalline topological materials the presence of boundary states in the insulating gap is not always necessary since they can be hidden in the bulk energy bands, obscured by boundary artifacts of non-topological or…
▽ More
The bulk-boundary correspondence, which links a bulk topological property of a material to the existence of robust boundary states, is a hallmark of topological insulators. However, in crystalline topological materials the presence of boundary states in the insulating gap is not always necessary since they can be hidden in the bulk energy bands, obscured by boundary artifacts of non-topological origin, or, in the case of higher-order topology, they can be gapped altogether. Crucially, in such systems the interplay between symmetry-protected topology and the corresponding symmetry defects can provide a variety of bulk probes to reveal their topological nature. For example, bulk crystallographic defects, such as disclinations and dislocations, have been shown to bind fractional charges and/or robust localized bound states in insulators protected by crystalline symmetries. Recently, exotic defects of translation symmetry called partial dislocations have been proposed as a probe of higher-order topology. However, it is a herculean task to have experimental control over the generation and probing of isolated defects in solid-state systems; hence their use as a bulk probe of topology faces many challenges. Instead, here we show that partial dislocation probes of higher-order topology are ideally suited to the context of engineered materials. Indeed, we present the first observations of partial-dislocation-induced topological modes in 2D and 3D higher-order topological insulators built from circuit-based resonator arrays. While rotational defects (disclinations) have previously been shown to indicate higher-order topology, our work provides the first experimental evidence that exotic translation defects (partial dislocations) are bulk topological probes.
△ Less
Submitted 3 May, 2021;
originally announced May 2021.
