"
• Stock Prediction: a method based on extraction of news features and recurrent neural networks abstract This paper proposed a method for stock prediction. In terms of feature extraction, we extract the features of stock-related news besides stock prices. We first select some seed words based on experience which are the symbols of good news and bad news. Then we propose an optimization method and calculate the positive polar of all words. After that, we construct the features of news based on the positive polar of their words. In consideration of sequential stock prices and continuous news effects, we propose a recurrent neural network model to help predict stock prices. Compared to SVM classifier with price features, we find our proposed method has an over 5% improvement on stock prediction accuracy in experiments.
• Ether: Bitcoin's competitor or ally? abstract Although Bitcoin has long been dominant in the crypto scene, it is certainly not alone. Ether is another cryptocurrency related project that has attracted an intensive attention because of its additional features. This study seeks to test whether these cryptocurrencies differ in terms of their volatile and speculative behaviors, hedge, safe haven and risk diversification properties. Using different econometric techniques, we show that a) Bitcoin and Ether are volatile and relatively more responsive to bad news, but the volatility of Ether is more persistent than that of Bitcoin; b) for both cryptocurrencies, the exuberance and the collapse of bubbles were identified, but Bitcoin appears more speculative than Ether; c) there is negative and significant correlation between Bitcoin/Ether and other assets (S\&amp;P500 stocks, US bonds, oil), which would indicate that digital currencies can hedge against the price movements of these assets; d) there is negative tail independence between Bitcoin/Ether and other financial assets, implying that these cryptocurrencies exhibit the function of a weak safe haven; and e) The inclusion of Bitcoin/ Ether in a portfolio improve its efficiency in terms of higher reward-to-risk ratios. But investors who hold diversified portfolios made of stocks or bonds and Ether may face losses over bearish regime. In such situation, stock and bond investors may take a short position on Bitcoin.
• Agent Inspired Trading Using Recurrent Reinforcement Learning and LSTM Neural Networks abstract With the breakthrough of computational power and deep neural networks, many areas that we haven't explore with various techniques that was researched rigorously in past is feasible. In this paper, we will walk through possible concepts to achieve robo-like trading or advising. In order to accomplish similar level of performance and generality, like a human trader, our agents learn for themselves to create successful strategies that lead to the human-level long-term rewards. The learning model is implemented in Long Short Term Memory (LSTM) recurrent structures with Reinforcement Learning or Evolution Strategies acting as agents The robustness and feasibility of the system is verified on GBPUSD trading.
• Extended Gini-type measures of risk and variability abstract The main of this paper is to introduce a family of risk measures which generalizes the Gini-type measures of risk and variability, by taking into consideration the psychological behavior. Our risk measures family is coherent and catches variability with respect to the decision-maker attitude towards risk.
• Optimal Trade Execution Under Endogenous Pressure to Liquidate: Theory and Numerical Solutions abstract We study optimal liquidation of a trading position (so-called block order or meta-order) in a market with a linear temporary price impact (Kyle, 1985). We endogenize the pressure to liquidate by introducing a downward drift in the unaffected asset price while simultaneously ruling out short sales. In this setting the liquidation time horizon becomes a stopping time determined endogenously, as part of the optimal strategy. We find that the optimal liquidation strategy is consistent with the square-root law which states that the average price impact per share is proportional to the square root of the size of the meta-order (Bershova and Rakhlin, 2013; Farmer et al., 2013; Donier et al., 2015; T\'oth et al., 2016).
Mathematically, the Hamilton-Jacobi-Bellman equation of our optimization leads to a severely singular and numerically unstable ordinary differential equation initial value problem. We provide careful analysis of related singular mixed boundary value problems and devise a numerically stable computation strategy by re-introducing time dimension into an otherwise time-homogeneous task.
• Statistical properties and multifractality of Bitcoin abstract Using 1-min high frequency returns of Bitcoin prices, we investigate statistical properties and multifractality of a Bitcoin time series. We find that the 1-min return distribution is fat-tailed and kurtosis largely deviates from the Gaussian expectation. Although with large time scales, kurtosis is anticipated to approach the Gaussian expectation, we find that convergence to that is very slow. Skewness is found to be negative at small time scales and becomes consistent with zero at large time scales. We also investigate daily volatility-asymmetry by using GARCH, GJR, and RGARCH models and find no evidence of volatility asymmetry. On exploring multifractality using multifractal detrended fluctuation analysis, we find that the Bitcoin time series exhibits multifractality. The sources of multifractality are also investigated and it is confirmed that both temporal correlation and the fat-tailed distribution contribute to the multifractality, and the degree of multifractality for the time correlation is stronger than that for the fat-tailed distribution.
• Lagrange regularisation approach to compare nested data sets and determine objectively financial bubbles' inceptions abstract Inspired by the question of identifying the start time $\tau$ of financial bubbles, we address the calibration of time series in which the inception of the latest regime of interest is unknown. By taking into account the tendency of a given model to overfit data, we introduce the Lagrange regularisation of the normalised sum of the squared residuals, $\chi^{2}_{np}(\Phi)$, to endogenously detect the optimal fitting window size := $w^* \in [\tau:\bar{t}_2]$ that should be used for calibration purposes for a fixed pseudo present time $\bar{t}_2$. The performance of the Lagrange regularisation of $\chi^{2}_{np}(\Phi)$ defined as $\chi^{2}_{\lambda (\Phi)}$ is exemplified on a simple Linear Regression problem with a change point and compared against the Residual Sum of Squares (RSS) := $\chi^{2}(\Phi)$ and RSS/(N-p):= $\chi^{2}_{np}(\Phi)$, where $N$ is the sample size and p is the number of degrees of freedom. Applied to synthetic models of financial bubbles with a well-defined transition regime and to a number of financial time series (US S\&amp;P500, Brazil IBovespa and China SSEC Indices), the Lagrange regularisation of $\chi^{2}_{\lambda}(\Phi)$ is found to provide well-defined reasonable determinations of the starting times for major bubbles such as the bubbles ending with the 1987 Black-Monday, the 2008 Sub-prime crisis and minor speculative bubbles on other Indexes, without any further exogenous information. It thus allows one to endogenise the determination of the beginning time of bubbles, a problem that had not received previously a systematic objective solution.
• Geopolitical Model of Investment Project Implementation abstract Two geopolitical actors implement a geopolitical project that involves transportaion and storage of some commodities. They interact with each other through a transport network. The network consists of several interconnected vertices. Some of the vetrices are trading hubs, storage spaces, production hubs and goods buyers. Actors wish to satify the demand of buyers and recieve the highest possible profit subject to compromise solution principle. A numerical example is given.
• The Correlation Structure of Anomaly Strategies
• Hybrid marked point processes: characterisation, existence and uniqueness abstract We introduce the class of hybrid marked point processes, which incorporate a state process that interacts with past-dependent events. For example, like in a Hawkes process, events can exhibit self- or cross-excitation effects, but these effects can now also depend on the state process. Events of type A will precipitate events of type B only when they move the state process to some critical region, say. In parallel, as each event occurs, the state process transitions to a new value according to transition probabilities that vary with the event type. We prove that such dynamics are equivalent to an intensity process of a specific product form. Our main result addresses the existence of non-explosive marked point processes with given intensities by studying a well-known Poisson-driven SDE (via Poisson embedding). The existing strong existence and uniqueness results rely on a Lipschitz condition that the intensity of a hybrid marked point process may fail to satisfy. This motivates us to propose a natural pathwise construction that instead requires only sublinear behaviour of the intensity. Using a domination argument, we are able to verify that this construction yields indeed a solution. As we restrict ourselves to non-explosive marked point processes, we also manage to prove uniqueness without any specific assumptions on the intensity.
• Breadth Momentum and Vigilant Asset Allocation (VAA): Winning More by Losing Less
• Second order stochastic differential models for financial markets abstract Using agent-based modelling, empirical evidence and physical ideas, such as the energy function and the fact that the phase space must have twice the dimension of the configuration space, we argue that the stochastic differential equations which describe the motion of financial prices with respect to real world probability measures should be of second order (and non-Markovian), instead of first order models \`a la Bachelier--Samuelson. Our theoretical result in stochastic dynamical systems shows that one cannot correctly reduce second order models to first order models by simply forgetting about momenta. We propose some simple second order models, including a stochastic constrained n-oscillator, which can explain many market phenomena, such as boom-bust cycles, stochastic quasi-periodic behavior, and "hot money" going from one market sector to another.
• Power-law tails in the distribution of order imbalance abstract We investigate the probability distribution of order imbalance calculated from the order flow data of 43 Chinese stocks traded on the Shenzhen Stock Exchange. Two definitions of order imbalance are considered based on the order number and the order size. We find that the order imbalance distributions of individual stocks have power-law tails. However, the tail index fluctuates remarkably from stock to stock. We also investigate the distributions of aggregated order imbalance of all stocks at different timescales $\Delta{t}$. We find no clear trend in the tail index with respect $\Delta{t}$. All the analyses suggest that the distributions of order imbalance are asymmetric.
• Contagious disruptions and complexity traps in economic development abstract Poor economies not only produce less; they typically produce things that involve fewer inputs and fewer intermediate steps. Yet the supply chains of poor countries face more frequent disruptions---delivery failures, faulty parts, delays, power outages, theft, government failures---that systematically thwart the production process. To understand how these disruptions affect economic development, we model an evolving input--output network in which disruptions spread contagiously among optimizing agents. The key finding is that a poverty trap can emerge: agents adapt to frequent disruptions by producing simpler, less valuable goods, yet disruptions persist. Growing out of poverty requires that agents invest in buffers to disruptions. These buffers rise and then fall as the economy produces more complex goods, a prediction consistent with global patterns of input inventories. Large jumps in economic complexity can backfire. This result suggests why "big push" policies can fail, and it underscores the importance of reliability and of gradual increases in technological complexity.
• Measuring the Knowledge Intensity of Economies with an Improved Measure of Economic Complexity abstract How much knowledge is there in an economy? In recent years, data on the mix of products that countries export has been used to construct measures of economic complexity that estimate the knowledge available in an economy and predict future economic growth. Here we introduce a new metric of economic complexity (ECI+) that measures the total exports of an economy corrected by how difficult it is to export each product. We use data from 1973 to 2013 to compare the ability of ECI+, the Economic Complexity Index (ECI), and Fitness complexity, to predict future economic growth using 5, 10, and 20-year panels in a pooled OLS, a random effects model, and a fixed effects model. We find that ECI+ outperforms ECI and Fitness in its ability to predict economic growth and in the consistency of its estimators across most econometric specifications. We then combine ECI+ with measures of physical capital, human capital, and institutions, to select a robust model of economic growth and test the robustness of ECI+. We find that the ability of ECI+ to predict growth is robust to these controls, and also, that human capital, political stability, and control of corruption, are positively associated with future economic growth, and initial level of income, is negatively associated with growth, in agreement with the traditional growth literature. Finally, we use ECI+ to generate economic growth predictions for the next 20 years and compare these predictions with the ones obtained using ECI and Fitness. These findings improve the methods available to estimate the knowledge intensity of economies using exports data and confirm the economic relevance of export structures.
• Linear and nonlinear correlations in order aggressiveness of Chinese stocks abstract The diagonal effect of orders is well documented in different markets, which states that orders are more likely to be followed by orders of the same aggressiveness and implies the presence of short-term correlations in order flows. Based on the order flow data of 43 Chinese stocks, we investigate if there are long-range correlations in the time series of order aggressiveness. The detrending moving average analysis shows that there are crossovers in the scaling behaviors of overall fluctuations and order aggressiveness exhibits linear long-term correlations. We design an objective procedure to determine the two Hurst indexes delimited by the crossover scale. We find no correlations in the short term and strong correlations in the long term for all stocks except for an outlier stock. The long-term correlation is found to depend on several firm specific characteristics. We also find that there are nonlinear long-term correlations in the order aggressiveness when we perform the multifractal detrending moving average analysis.
• Wisdom of the Employee Crowd: Employer Reviews and Stock Returns
• Wax and wane of the cross-sectional momentum and contrarian effects: Evidence from the Chinese stock markets abstract This paper investigates the time-varying risk-premium relation of the Chinese stock markets within the framework of cross-sectional momentum and contrarian effects by adopting the Capital Asset Pricing Model and the French-Fama three factor model. The evolving arbitrage opportunities are also studied by quantifying the performance of time-varying cross-sectional momentum and contrarian effects in the Chinese stock markets. The relation between the contrarian profitability and market condition factors that could characterize the investment context is also investigated. The results reveal that the risk-premium relation varies over time, and the arbitrage opportunities based on the contrarian portfolios wax and wane over time. The performance of contrarian portfolios are highly dependent on several market conditions. The periods with upward trend of market state, higher market volatility and liquidity, lower macroeconomics uncertainty are related to higher contrarian profitability. These findings are consistent with the Adaptive Markets Hypothesis and have practical implications for market participants.
• A Behavioral Modification of Popular Option Pricing Models
• Investing with Cryptocurrencies - A Liquidity Constrained Investment Approach