ORiON

Editorial

2024-07-01T21:05:46+00:00

Assessing the Performance of Machine Learning Models for Default Prediction under Missing Data and Class Imbalance: A Simulation Study

2024-07-03T11:53:46+00:00

In the field of machine learning, robust model performance is essential for accurate predictions and informed decision-making. One critical challenge that hampers the effectiveness of machine learning algorithms is the presence of missing data. Missing values are ubiquitous in real-world datasets and can significantly impact the performance of predictive models. This study explores the impact of increasing levels of missing values on the performance of machine learning models. Simulated samples with missing values ranging from 5% to 50% were generated, and various models were evaluated accordingly. Missing data is a prevalent change that hinders the performance of machine learning algorithms. The results demonstrated a consistent trend of deteriorating model performance as the amount of missing values increases. Higher levels of missing values lead to decreased accuracy scores across all models. Among the models evaluated, decision trees (DT) and random forests (RF) consistently demonstrated high accuracy scores across all sampling techniques, showcasing their robustness in handling missing values. Logistic regression (LR) also performed relatively well, showing consistent performance across different levels of missing values. On the other hand, stochastic gradient descent classifier (SGDC), K-nearest neighbors (kNN), and naive Bayes (NB) models consistently exhibited lower accuracy scores across all sampling techniques, indicating limitations in handling missing values even when the dataset was more balanced. Furthermore, the study highlights the superiority of the SMOTE (Synthetic Minority OVER-sampling Technique) sampling technique compared to the UNDER-sampling approach. Models trained using SMOTE consistently achieved higher accuracy scores across all levels of missing values. This suggests that SMOTE sampling effectively handles imbalanced datasets and enhances classification performance, particularly when dealing with missing values. In an era where data fuels decision-making, this study's insights into the escalating impact of missing values on machine learning models stand as a clarion call for robust data handling techniques. As the quest for accurate predictions gains paramount importance, addressing the pervasive challenge of missing data emerges as a cornerstone for unlocking the true potential of machine learning in real-world applications.

Static hedging of vanilla and exotic options in a South African context

2024-07-01T21:05:46+00:00

In this paper, we test the performance of a static hedging strategy for a long-dated European call option and European spread call option in South Africa. The stochastic volatility double jump (SVJJ) model is calibrated to historical FTSE/JSE Top40 returns to generate real-world FTSE/JSE Top40 prices at future dates. The SVJJ model is also calibrated to the FTSE/JSE (Top40) implied volatility surface in order to value the options under the risk-neutral measure. Two static hedging programs are then implemented to test their effectiveness when replicating a long-dated European call option and European spread call option. Our results indicate that static hedging is a simple, yet effective, solution when hedging non-exchange-traded options with vanilla exchange-traded options.

The profitability of flat-price broadband with an over-the-top subscription content product – benefits from cooperation

2024-07-01T21:05:46+00:00

We investigate the effect on the broadband market when an over-the-top subscription content product is introduced. Does it necessarily increase or decrease profitability of the broadband product when it (a) boosts the utility of broadband but (b) imposes additional costs to deliver the broadband service?

The short answer is it depends. The different scenarios that we choose to illustrate this demonstrate that in many cases, the broadband and content providers can jointly benefit from coordination on how the content product is priced. Empirical evidence confirms that coordination does take place where 'network neutrality' is not mandated. In addition to the illustrative scenarios, we run a large number of simulations with a single broadband and single content provider, restricting the firms to integer prices, for different distributions of the customer valuations.

The results show that cooperation between the firms (possibly through paid peering) generally produces better outcomes (also from the consumer point of view) than when the broadband provider reacts by raising the price of broadband independently.

An optimisation approach towards soccer Fantasy Premiere League team selection

2024-07-01T21:05:46+00:00

\textit{Fantasy Premiere League} (FPL) is a popular online sports prediction game based on the well-known (soccer) \textit{English Premiere League} (EPL). In FPL, each participant forms a series of imaginary teams over the course of a season of the EPL, each composed of real-world soccer players. Points are then awarded to participants based on the real-world performances of the players in the EPL. The goal is to accumulate as many points as possible during the season. FPL Participants are allocated a budget for team selection which gives rise to the constrained team selection optimisation problem of filling playing positions in the team for each game week of the FPL season. This team selection problem is further complicated by the facts that player performance is not known in advance with certainty and that participants may only make limited changes to their team compositions in the form of transfers during any game week. In this paper we adopt a combinatorial optimisation approach towards participating in FPL, in which future player performances are forecast by statistical and machine learning techniques. We demonstrate retrospectively that our approach would have placed within the top 4\% of players worldwide during the 2020/2021 FPL season.