Browsing by Subject "Calibration"

Now showing 1 - 3 of 3
  • Thumbnail Image
    Item
    Open Access
    The LOFAR Two-metre Sky Survey: I. Survey description and preliminary data release⋆
    (2017) Shimwell, T W; Röttgering, H J A; Best, P N; Williams, W L; Dijkema, T J; de Gasperin, F; Hardcastle, M J; Heald, G H; Hoang, D N; Horneffer, A; Intema, H; Mahony, E K; Mandal, S; Mechev, A P; Morabito, L; Oonk, J B R; Rafferty, D; Retana-Montenegro, E; Sabater, J; Tasse, C; van Weeren, R J; BrYggen, M; Brunetti, G; Chyży, K T; Conway, J E; Haverkorn, M; Jackson, N; Jarvis, M J; McKean, J P; Miley, G K; Morganti, R; White, G J
    598
  • Thumbnail Image
    Item
    Open Access
    Validation of two prediction models of undiagnosed chronic kidney disease in mixed-ancestry South Africans
    (BioMed Central Ltd, 2015) Mogueo, Amelie; Echouffo-Tcheugui, Justin; Matsha, Tandi; Erasmus, Rajiv; Kengne, Andre
    BACKGROUND: Chronic kidney disease (CKD) is a global challenge. Risk models to predict prevalent undiagnosed CKD have been published. However, none was developed or validated in an African population. We validated the Korean and Thai CKD prediction model in mixed-ancestry South Africans. METHODS: Discrimination and calibration were assessed overall and by major subgroups. CKD was defined as 'estimated glomerular filtration rate (eGFR) <60ml/min/1.73m 2 ' or 'any nephropathy'. eGFR was based on the 4-variable Modification of Diet in Renal Disease (MDRD) formula. RESULTS: In all 902 participants (mean age 55years) included, 259 (28.7%) had prevalent undiagnosed CKD. C-statistics were 0.76 (95 % CI: 0.73-0.79) for 'eGFR <60ml/min/1.73m 2 ' and 0.81 (0.78-0.84) for 'any nephropathy' for the Korean model; corresponding values for the Thai model were 0.80 (0.77-0.83) and 0.77 (0.74-0.81). Discrimination was better in men, older and normal weight individuals. The model underestimated CKD risk by 10% to 13% for the Thai and 9% to 93% for the Korean model. Intercept adjustment significantly improved the calibration with an expected/observed risk of 'eGFR <60ml/min/1.73m 2 ' and 'any nephropathy' respectively of 0.98 (0.87-1.10) and 0.97 (0.86-1.09) for the Thai model; but resulted in an underestimation by 24% with the Korean model. Results were broadly similar for CKD derived from the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) formula. CONCLUSION: Asian prevalent CKD risk models had acceptable performances in mixed-ancestry South Africans. This highlights the potential importance of using existing models for risk CKD screening in developing countries.
  • Thumbnail Image
    Item
    Open Access
    Volatility Model Pricing and Calibration with Neural Networks using Bayesian Optimisation
    (2022) Goosen, Jenna; Ouwehand, Peter
    Stochastic Alpha, Beta, Rho (SABR) and Heston Volatility models have been used in the financial industry due to their ability to price options as a function of time to maturity and moneyness. Implied volatilities for these models are accurately estimated using a numerical integration approach, Monte Carlo approach, series expansion approximation or using a two factor finite difference approach. However, these volatility calculations are computationally expensive. Therefore, this dissertation explores the use of deep artificial neural networks to calibrate volatility models by deploying computational resources to train a model (offline) and then utilise the pre-trained model to competitively price options (online). Deep neural networks in this paper are trained using an indirect and a direct method. The first step of the indirect method utilises a deep artificial neural network to approximate the pricing function (output) using the parameters as inputs. The second step of the indirect method implements a least squares optimisation algorithm to calibrate the parameters. The direct method, on the other hand, uses a deep artificial neural network to calibrate the model parameters using the implied volatility surface as input. Bayesian Optimisation algorithms are implemented to select models with the lowest loss metric for SABR and Heston volatility models. The best performing models are tested and compared for accuracy, speed and robustness for pricing and calibration. This dissertation finds that deep artificial neural networks using Bayesian Optimisation for the indirect and direct methods are able to efficiently and accurately price and calibrate the SABR and Heston Model. In addition, the results show that the implementation of the indirect and direct method, when there is no closed form approximation readily available, is advantageous from both a time and accuracy perspective for pricing and calibration.