The combination of neural estimates in prediction and decision problems

In this dissertation, different ways of combining neural predictive models or neural-based forecasts are discussed. The proposed approaches consider mostly Gaussian radial basis function networks, which can be efficiently identified and estimated through recursive/adaptive methods. Two different ways of combining are explored to get a final estimate – model mixing and model synthesis –, with the aim of obtaining improvements both in terms of efficiency and effectiveness. In the context of model mixing, the usual framework for linearly combining estimates from different models is extended, to deal with the case where the forecast errors from those models are correlated. In the context of model synthesis, and to address the problems raised by heavily nonstationary time series, we propose hybrid dynamic models for more advanced time series forecasting, composed of a dynamic trend regressive model (or, even, a dynamic harmonic regressive model), and a Gaussian radial basis function network. Additionally, using the model mixing procedure, two approaches for decision-making from forecasting models are discussed and compared: either inferring decisions from combined predictive estimates, or combining prescriptive solutions derived from different forecasting models. Finally, the application of some of the models and methods proposed previously is illustrated with two case studies, based on time series from finance and from tourism.

Development of a dynamic headspace solid-phase microextraction procedure coupled to GC–qMSD for evaluation the chemical profile in alcoholic beverages

In the present study, a simple and sensitive methodology based on dynamic headspace solid-phase microextraction (HS-SPME) followed by thermal desorption gas chromatography with quadrupole mass detection (GC–qMSD), was developed and optimized for the determination of volatile (VOCs) and semi-volatile (SVOCs) compounds from different alcoholic beverages: wine, beer and whisky. Key experimental factors influencing the equilibrium of the VOCs and SVOCs between the sample and the SPME fibre, as the type of fibre coating, extraction time and temperature, sample stirring and ionic strength, were optimized. The performance of five commercially available SPME fibres was evaluated and compared, namely polydimethylsiloxane (PDMS, 100 μm); polyacrylate (PA, 85 μm); polydimethylsiloxane/divinylbenzene (PDMS/DVB, 65 μm); carboxen™/polydimethylsiloxane (CAR/PDMS, 75 μm) and the divinylbenzene/carboxen on polydimethylsiloxane (DVB/CAR/PDMS, 50/30 μm) (StableFlex). An objective comparison among different alcoholic beverages has been established in terms of qualitative and semi-quantitative differences on volatile and semi-volatile compounds. These compounds belong to several chemical families, including higher alcohols, ethyl esters, fatty acids, higher alcohol acetates, isoamyl esters, carbonyl compounds, furanic compounds, terpenoids, C13-norisoprenoids and volatile phenols. The optimized extraction conditions and GC–qMSD, lead to the successful identification of 44 compounds in white wines, 64 in beers and 104 in whiskys. Some of these compounds were found in all of the examined beverage samples. The main components of the HS-SPME found in white wines were ethyl octanoate (46.9%), ethyl decanoate (30.3%), ethyl 9-decenoate (10.7%), ethyl hexanoate (3.1%), and isoamyl octanoate (2.7%). As for beers, the major compounds were isoamyl alcohol (11.5%), ethyl octanoate (9.1%), isoamyl acetate (8.2%), 2-ethyl-1-hexanol (5.9%), and octanoic acid (5.5%). Ethyl decanoate (58.0%), ethyl octanoate (15.1%), ethyl dodecanoate (13.9%) followed by 3-methyl-1-butanol (1.8%) and isoamyl acetate (1.4%) were found to be the major VOCs in whisky samples.

Simultaneous analysis of free amino acids and biogenic amines in honey and wine samples using in loop orthophthalaldeyde derivatization procedure

This work presents a RP-HPLC method for the simultaneous quantification of free amino acids and biogenic amines in liquid food matrices and the results of the application to honey and wine samples obtained from different production processes and geographic origins. The developed methodology is based on a pre-column derivatization with o-phthaldialdehyde carried out in the sample injection loop. The compounds were separated in a Nova-Pack RP-C18 column (150 mm × 3.9 mm, 4 μm) at 35 °C. The mobile phase used was a mixture of phase A: 10 mM sodium phosphate buffer (pH 7.3), methanol and tetrahydrofuran (91:8:1); and phase B: methanol and phosphate buffer (80:20), with a flow rate of 1.0 ml/min. Fluorescence detection was used at an excitation wavelength of 335 nm and an emission wavelength of 440 nm. The separation and quantification of 19 amino acids and 6 amines was carried out in a single run as their OPA/MCE derivatives elute within 80 min, ensuring a reproducible quantification. The method showed to be adequate for the purpose, with an average RSD of 2% for the different amino acids; detection limits varying between 0.71 mg/l (Asn) and 8.26 mg/l (Lys) and recovery rates between 63.0% (Cad) and 98.0% (Asp). The amino acids present at the highest concentration in honey and wine samples were phenylalanine and arginine, respectively. Only residual levels of biogenic amines were detected in the analysed samples.