Tiejun Ma

Bridging the divide in financial market forecasting

An extensive benchmark in financial time series forecasting is performed.Best machine learning(ML) methods out-perform best econometric methods.The ML methodology employed significantly affects forecasting accuracy.Market maturity, forecast horizon & model-assessment method affect forecast accuracy.Evidence against the informational value of technical indicators. Financial time series forecasting is a popular application of machine learning methods. Previous studies report that advanced forecasting methods predict price changes in financial markets with high accuracy and that profit can be made trading on these predictions. However, financial economists point to the informational efficiency of financial markets, which questions price predictability and opportunities for profitable trading. The objective of the paper is to resolve this contradiction. To this end, we undertake an extensive forecasting simulation, based on data from thirty-four financial indices over six years. These simulations confirm that the best machine learning methods produce more accurate forecasts than the best econometric methods. We also examine the methodological factors that impact the predictive accuracy of machine learning forecasting experiments. The results suggest that the predictability of a financial market and the feasibility of profitable model-based trading are significantly influenced by the maturity of the market, the forecasting method employed, the horizon for which it generates predictions and the methodology used to assess the model and simulate model-based trading. We also find evidence against the informational value of indicators from the field of technical analysis. Overall, we confirm that advanced forecasting methods can be used to predict price changes in some financial markets and we discuss whether these results question the prevailing view in the financial economics literature that financial markets are efficient.

A new methodology for generating and combining statistical forecasting models to enhance competitive event prediction

Forecasting methods are routinely employed to predict the outcome of competitive events (CEs) and to shed light on the factors that influence participants’ winning prospects (e.g., in sports events, political elections). Combining statistical models’ forecasts, shown to be highly successful in other settings, has been neglected in CE prediction. Two particular difficulties arise when developing model-based composite forecasts of CE outcomes: the intensity of rivalry among contestants, and the strength/diversity trade-off among individual models. To overcome these challenges we propose a range of surrogate measures of event outcome to construct a heterogeneous set of base forecasts. To effectively extract the complementary information concealed within these predictions, we develop a novel pooling mechanism which accounts for competition among contestants: a stacking paradigm integrating conditional logit regression and log-likelihood-ratio-based forecast selection. Empirical results using data related to horseracing events demonstrate that: (i) base model strength and diversity are important when combining model-based predictions for CEs; (ii) average-based pooling, commonly employed elsewhere, may not be appropriate for CEs (because average-based pooling exclusively focuses on strength); and (iii) the proposed stacking ensemble provides statistically and economically accurate forecasts. These results have important implications for regulators of betting markets associated with CEs and in particular for the accurate assessment of market efficiency.

Rule-Based Verification of Network Protocol Implementations Using Symbolic Execution

The secure and correct implementation of network protocols for resource discovery, device configuration and network management is complex and error-prone. Protocol specifications contain ambiguities, leading to implementation flaws and security vulnerabilities in network daemons. Such problems are hard to detect because they are often triggered by complex sequences of packets that occur only after prolonged operation. The goal of this work is to find semantic bugs in network daemons. Our approach is to replay a set of input packets that result in high source code coverage of the daemon and observe potential violations of rules derived from the protocol specification. We describe SYMNV, a practical verification tool that first symbolically executes a network daemon to generate high coverage input packets and then checks a set of rules constraining permitted input and output packets. We have applied SYMNV to three different implementations of the Zeroconf protocol and show that it is able to discover non-trivial bugs.

On the Quality of Service of Crash-Recovery Failure Detectors

In this paper, we study and model a crash-recovery target and its failure detectors probabilistic behavior. We extend quality of service (QoS) metrics to measure the recovery detection speed and the proportion of the detected failures of a crash-recovery failure detector. Then the impact of the dependability of the crash-recovery target on the QoS bounds for such a crash-recovery failure detector is analysed by adopting general dependability metrics such as MTTF and MTTR. In addition, we analyse how to estimate the failure detectors parameters to achieve the QoS from a requirement based on Chens NFD-S algorithm. We also demonstrate how to execute the configuration procedure of this crash-recovery failure detector. The simulations are based on the revised NFD-S algorithm with various MTTF and MTTR. The simulation results show that the dependability of a recoverable monitored target could have significant impact on the QoS of such a failure detector and match our analysis results.

Towards automated verification of autonomous networks: A case study in self-configuration

In autonomic networks, the self-configuration of network entities is one of the most desirable properties. In this paper, we show how formal verification techniques can verify the correctness of self-configuration. As a case study, we describe the configuration of physical cell identifiers (PCIs), a radio configuration parameter in cellular base stations. We provide formal models of PCI assignment algorithms and their desired properties. We then demonstrate how the potential for conflicting PCI assignments can be detected using model checking and resolved in the design stage. Through this case study, we argue that both simulation and verification should be adopted and highlight the potential of runtime verification approaches in this space.

Evaluation of the QoS of crash-recovery failure detection

Crash failure detection is a key topic in fault tolerance, and it is important to be able to assess the QoS of failure detection services. Most previous work on crash failure detectors has been based on the crash-stop or fail-free assumption. In this paper we study and model a crash-recovery service which has the ability to recover from the crash state. We analyse the QoS bounds for such a crash-recovery failure detection service. Our results show that the dependability metrics of the monitored service will have an impact on the QoS of the failure detection service. Our results are corroborated by simulation results, showing bounds on the QoS.

Social Machines: how recent technological advances have aided financialisation

In recent years, financial markets have been fundamentally transformed by innovations in information technology, in particular with regard to the web, social networks, high-speed computer networks and mobile technologies. We borrow the concept of Social Machines from Web Science as a single concept that captures the essence of all these recent technological changes to argue that the emergence of these Social Machines has aided the transformation of financial markets and society. This study explores the formation of these Social Machines with three sample disruptive technologies – automated/high-frequency trading, social network analytics and smart mobile technology. Through critical reflective analysis of these three case studies, we assess the impact of information technology innovation on financialisation. We adopt three case studies – automated trading; market information extraction using social media technologies; and information diffusion and trader decision-making with mobile technology on financial and real sector changes – which demonstrate the increasing trend of transaction velocity, speculative trading, increased complex information network, accelerated inequality and leverage. Our findings demonstrate that technologically enabled financial Social Machines harness crowd wisdom, engage disparate individual traders to produce more accurate price estimations, and have enhanced decision-making capability. However, these same changes can also have a simultaneously detrimental effect on financial and real sectors, in some situations exacerbating underlying distortions, such as misinformation due to complex information networks, speculative trading behaviour, and higher volatility with transaction velocity. Overall, we conclude that these innovations have transformed the fundamental nature of key aspects of the finance industry and society as a whole.

Time is money: Costing the impact of duration misperception in market prices

We explore whether, and to what extent, traders in a real world financial market, where participants’ judgements are reportedly well calibrated, are subject to duration misperception. To achieve this, we examine duration misperception in the horserace betting market. We develop a two-stage algorithm to predict horses’ winning probabilities that account for a duration-related factor that is known to affect horses’ winning prospects. The algorithm adapts survival analysis and combines it with the conditional logit model. Using a dataset of 4736 horseraces and the lifetime career statistics of the 53,295 horses running in these races, we demonstrate that prices fail to discount fully information related to duration since a horses last win. We show that this failure is extremely costly, since a betting strategy based on the predictions arising from the model shows substantial profits (932.5 percent and 16.27 percent, with and without reinvestment of winnings, respectively). We discuss the important implications of duration neglect in the wider economy.

Combining POS Tagging, Lucene Search and Similarity Metrics for Entity Linking

Entity linking is to detect proper nouns or concrete concepts (a.k.a mentions) from documents, and to map them to the corresponding entries in a given knowledge base. In this paper, we propose an entity linking framework POSLS consisting of three components: mention detection, candidate selection and entity disambiguation. First, we use part of speech tagging and English syntactic rules to detect mentions. We then choose candidates with Lucene search. Finally, we identify the best matchings with a similarity based disambiguation method. Experimental results show that our approach has an acceptable accuracy.

Femtocell coverage optimisation using statistical verification

Femtocells are small base stations that provide radio coverage for mobile devices in homes or office areas. In this paper, we consider the optimisation of a number of femtocells that provide joint coverage in enterprise environments. In such an environment, femtocells should minimise coverage overlap and coverage holes and ensure a balanced traffic workload among them. We use statistical verification techniques to monitor the probabilistic correctness of a given femtocell configuration at runtime. If there is any violation of the desired level of service, a self-optimisation procedure is triggered to improve the current configuration. Our evaluation results show that, compared with fixed time, interval-based optimisation, our approach achieves better coverage and can detect goal violations quickly with a given level of confidence when they occur frequently. It can also avoid unnecessary self-optimisation cycles, reducing the cost of self-optimisation.