Quantitative Finance

In this paper we propose a deep recurrent model based on the order flow for the stationary modelling of the high-frequency directional prices movements. The order flow is the microsecond stream of orders arriving at the exchange, driving the formation of prices seen on the price chart of a stock or currency. To test the stationarity of our proposed model we train our model on data before the 2017 Bitcoin bubble period and test our model during and after the bubble. We show that without any retraining, the proposed model is temporally stable even as Bitcoin trading shifts into an extremely volatile "bubble trouble" period. The significance of the result is shown by benchmarking against existing state-of-the-art models in the literature for modelling price formation using deep learning.

Prediction of stock groups' values has always been attractive and challenging for shareholders. This paper concentrates on the future prediction of stock market groups. Four groups named diversified financials, petroleum, non-metallic minerals and basic metals from Tehran stock exchange are chosen for experimental evaluations. Data are collected for the groups based on ten years of historical records. The values predictions are created for 1, 2, 5, 10, 15, 20 and 30 days in advance. The machine learning algorithms utilized for prediction of future values of stock market groups. We employed Decision Tree, Bagging, Random Forest, Adaptive Boosting (Adaboost), Gradient Boosting and eXtreme Gradient Boosting (XGBoost), and Artificial neural network (ANN), Recurrent Neural Network (RNN) and Long short-term memory (LSTM). Ten technical indicators are selected as the inputs into each of the prediction models. Finally, the result of predictions is presented for each technique based on three metrics. Among all the algorithms used in this paper, LSTM shows more accurate results with the highest model fitting ability. Also, for tree-based models, there is often an intense competition between Adaboost, Gradient Boosting, and XGBoost.

We introduce a non linear pricing model of individual stock returns that defines a stickiness parameter of the returns. The pricing model resembles the capital asset pricing model used in finance but has a non linear component inspired from models of earth quake tectonic plate movements. The link to tectonic plate movements happens, since price movements of a given stock index is seen adding stress to its components of individual stock returns, in order to follow the index. How closely individual stocks follow the indexs price movements, can then be used to define their stickiness

In the paper, the martingales and super-martingales relative to a regular set of measures are systematically studied. The notion of local regular super-martingale relative to a set of equivalent measures is introduced and the necessary and sufficient conditions of the local regularity of it in the discrete case are founded. The regular set of measures play fundamental role for the description of incomplete markets. In the partial case, the description of the regular set of measures is presented. The notion of completeness of the regular set of measures have the important significance for the simplification of the proof of the optional decomposition for super-martingales. Using this notion, the important inequalities for some random values are obtained. These inequalities give the simple proof of the optional decomposition of the majorized super-martingales. The description of all local regular super-martingales relative to the regular set of measures is presented. It is proved that every majorized super-martingale relative to the complete set of measures is a local regular one. In the case, as evolution of a risk asset is given by the discrete geometric Brownian motion, the financial market is incomplete and a new formula for the fair price of super-hedge is founded.

We present a detailed study of the performance of a trading rule that uses moving average of past returns to predict future returns on stock indexes. Our main goal is to link performance and the stochastic process of the traded asset. Our study reports short, medium and long term effects by looking at the Sharpe ratio (SR). We calculate the Sharpe ratio of our trading rule as a function of the probability distribution function of the underlying traded asset and compare it with data. We show that if the performance is mainly due to presence of autocorrelation in the returns of the traded assets, the SR as a function of the portfolio formation period (look-back) is very different from performance due to the drift (average return). The SR shows that for look-back periods of a few months the investor is more likely to tap into autocorrelation. However, for look-back larger than few months, the drift of the asset becomes progressively more important. Finally, our empirical work reports a new long-term effect, namely oscillation of the SR and propose a non-stationary model to account for such oscillations.

Detecting changes in asset co-movements is of much importance to financial practitioners, with numerous risk management benefits arising from the timely detection of breakdowns in historical correlations. In this article, we propose a real-time indicator to detect temporary increases in asset co-movements, the Autoencoder Reconstruction Ratio, which measures how well a basket of asset returns can be modelled using a lower-dimensional set of latent variables. The ARR uses a deep sparse denoising autoencoder to perform the dimensionality reduction on the returns vector, which replaces the PCA approach of the standard Absorption Ratio, and provides a better model for non-Gaussian returns. Through a systemic risk application on forecasting on the CRSP US Total Market Index, we show that lower ARR values coincide with higher volatility and larger drawdowns, indicating that increased asset co-movement does correspond with periods of market weakness. We also demonstrate that short-term (i.e. 5-min and 1-hour) predictors for realised volatility and market crashes can be improved by including additional ARR inputs.

Multifractal detrended cross-correlation methodology is described and applied to Foreign exchange (Forex) market time series. Fluctuations of high frequency exchange rates of eight major world currencies over 2010-2018 period are used to study cross-correlations. The study is motivated by fundamental questions in complex systems' response to significant environmental changes and by potential applications in investment strategies, including detecting triangular arbitrage opportunities. Dominant multiscale cross-correlations between the exchange rates are found to typically occur at smaller fluctuation levels. However hierarchical organization of ties expressed in terms of dendrograms, with a novel application of the multiscale cross-correlation coefficient, are more pronounced at large fluctuations. The cross-correlations are quantified to be stronger on average between those exchange rate pairs that are bound within triangular relations. Some pairs from outside triangular relations are however identified to be exceptionally strongly correlated as compared to the average strength of triangular correlations.This in particular applies to those exchange rates that involve Australian and New Zealand dollars and reflects their economic relations. Significant events with impact on the Forex are shown to induce triangular arbitrage opportunities which at the same time reduce cross--correlations on the smallest time scales and act destructively on the multiscale organization of correlations. In 2010--2018 such instances took place in connection with the Swiss National Bank intervention and the weakening of British pound sterling accompanying the initiation of Brexit procedure. The methodology could be applicable to temporal and multiscale pattern detection in any time series.

We present an advance bubble detection methodology based on the Log Periodic Power Law Singularity (LPPLS) confidence indicator for the early causal identification of positive and negative bubbles in the Chinese stock market using the daily data on the Shanghai Shenzhen CSI 300 stock market index from January 2002 through April 2018. We account for the damping condition of LPPLS model in the search space and implement the stricter filter conditions for the qualification of the valid LPPLS fits by taking account of the maximum relative error, performing the Lomb log-periodic test of the detrended residual, and unit-root tests of the logarithmic residual based on both the Phillips-Perron test and Dickey-Fuller test to improve the performance of LPPLS confidence indicator. Our analysis shows that the LPPLS detection strategy diagnoses the positive bubbles and negative bubbles corresponding to well-known historical events, implying the detection strategy based on the LPPLS confidence indicator has an outstanding performance to identify the bubbles in advance. We find that the probability density distribution of the estimated beginning time of bubbles appears to be skewed and the mass of the distribution is concentrated on the area where the price starts to have an obvious super-exponentially growth. This study is the first work in the literature that identifies the existence of bubbles in the Chinese stock market using the daily data of CSI 300 index with the advance bubble detection methodology of LPPLS confidence indicator. We have shown that it is possible to detect the potential positive and negative bubbles and crashes ahead of time, which in turn limits the bubble sizes and eventually minimizes the damages from the bubble crash.

We determine the number of statistically significant factors in a forecast model using a random matrices test. The applied forecast model is of the type of Reduced Rank Regression (RRR), in particular, we chose a flavor which can be seen as the Canonical Correlation Analysis (CCA). As empirical data, we use cryptocurrencies at hour frequency, where the variable selection was made by a criterion from information theory. The results are consistent with the usual visual inspection, with the advantage that the subjective element is avoided. Furthermore, the computational cost is minimal compared to the cross-validation approach.

The enlargement of the European Union to new countries in 2004 launched mechanisms supporting the development of various social and economic areas, as well as levelling the differences between the Community members in these areas. This article focuses on the insurance sector. Its main purpose is to analyze the development and similarity of the insurance markets of old and new members of the European Union after the enlargement in 2004.