Market prices are traditionally recorded in fixed time intervals. Directional Change is an alternative approach to summarize price movements in financial markets that is consistent with across all time scales. Unlike time series, directional change summarizes the big data in finance by focusing on the intrinsic time of the data. This captures deeper intrinsic data qualities and thus trading strategies based on directional change are more sustainable and less disruptive. In this paper, we propose four trading strategies using the concept of directional change and explore the combination with technical analysis. The trading strategies are tested using EUR/USD and GBP/USD high frequency FX market data. Empirical results show good performance of our trading strategies based on thresholds, and that combining with technical analysis brings further improvement.

A financial asset's volatility exhibits key characteristics, such as mean-reversion and high autocorrelation [1], [2]. Empirical evidence suggests that this volatility autocorrelation exponentially decays (or exhibits long-range memory) [3]. We employ Genetic Programming (GP) for volatility forecasting because of its ability to detect patterns such as the conditional mean and conditional variance of a time-series. Genetic Programming is typically applied to optimisation, searching, and machine learning applications like classification, prediction etc. From our experiments, we see that Genetic Programming is a good competitor to the standard forecasting techniques like GARCH(1,1), Moving Average (MA), Exponentially Weighted Moving Average (EWMA). However it is not a silver bullet: we observe that different forecasting methods would perform better in different market conditions. In addition to Genetic Programming, we consider a heuristic technique that employs a series of standard forecasting methods and dynamically opts for the most appropriate technique at a given time. Using a heuristic technique, we try to identify the best forecasting method that would perform better than the rest of the methods in the near out-of-sample horizon. Our work introduces a preliminary framework for forecasting 5-day annualised volatility in GBP/USD, USD/JPY, and EUR/USD.

In this paper, we use reinforcement learning (RL) as a tool to study price dynamics in an electronic retail market consisting of two competing sellers, and price sensitive and lead time sensitive customers. Sellers, offering identical products, compete on price to satisfy stochastically arriving demands (customers), and follow standard inventory control and replenishment policies to manage their inventories. In such a generalized setting, RL techniques have not previously been applied. We consider two representative cases: 1) no information case, were none of the sellers has any information about customer queue levels, inventory levels, or prices at the competitors; and 2) partial information case, where every seller has information about the customer queue levels and inventory levels of the competitors. Sellers employ automated pricing agents, or pricebots, which use RL-based pricing algorithms to reset the prices at random intervals based on factors such as number of back orders, inventory levels, and replenishment lead times, with the objective of maximizing discounted cumulative profit. In the no information case, we show that a seller who uses Q-learning outperforms a seller who uses derivative following (DF). In the partial information case, we model the problem as a Markovian game and use actor-critic based RL to learn dynamic prices. We believe our approach to solving these problems is a new and promising way of setting dynamic prices in multiseller environments with stochastic demands, price sensitive customers, and inventory replenishments.

Procurement, the process of obtaining materials or services, is a critical process for any organization. While procuring a set of items from different suppliers who may sell only a subset (bundle) of a desired set of items, it will be required to select an optimal set of suppliers who can supply the desired set of items. This is the optimal vendor selection problem. Bundling in procurement has benefits such as demand aggregation, supplier aggregation, and lead-time reduction. The NP-hardness of the vendor selection problem motivates us to formulate a compatible linear programming problem by relaxing the integer constraints and imposing additional constraints. The newly formulated problem can be solved by a novel iterative algorithm proposed recently in the literature. In this paper, we show that the application of this iterative algorithm leads to an iterative procurement auction that improves the efficiency of the procurement process. By using reinforcement learning to orchestrate the iterations of the algorithm, we show impressive gains in computational efficiency of the algorithm.

Procurement, i.e., the process of obtaining materials or services and managing their inflow into organizations, is a critical process in supply chain management. Internet technologies offer attractive opportunities to reduce purchasing costs while increasing the speed and efficiency of the procurement process through seamless network integration and automated execution of key tasks. The paper attempts to model e-procurement from a lead time perspective, to capture the effects of Internet technologies on the procurement lead time. The model we develop is a multiclass queueing network called a probabilistic reentrant line. The model captures several important facets of an e-procurement process, such as use of electronic RFQ (request for quotation) in systematic sourcing, use of Internet auctions in spot purchasing, contention for procurement resources, randomness in execution times of procurement tasks, and concurrent execution of multiple procurement activities. The analysis of the queueing model suggests that end-to-end lead times in e-procurement can be compressed in ingenious ways using standard queueing theoretic techniques.