预测造纸废水出水指标的随机森林建模方法

出水化学需氧量(COD)与出水固形物含量(SS)是评价造纸废水处理工艺好坏的重要指标。为了更好地对其进行预测,提出了一种基于随机森林(RF)模型的方法,并以R语言为工具进行回归预测。对比偏最小二乘(PLS)模型、支持向量回归(SVR)模型、人工神经网络(ANN)模型等常规预测模型,发现RF模型具有预测精度高,结果误差小,泛化能力好,调整参数少等优点。在对出水COD进行预测时,RF模型的相关系数r为0.7954,相比于PLS、SVR、ANN分别提高了8.88%、10.73%、14.68%。在对出水SS进行预测时,RF模型的相关系数r为0.8551,相比于PLS、SVR、ANN分别提高了15.43%、24.25%、30.79%。     

西式快餐用煎炸油质量近红外快速检测模型的建立

为对西式快餐用煎炸油质量进行快速监测,选用近红外光谱法（NIRS）联合偏最小二乘法（PLS）,分别建立酸价和极性组分两个煎炸油质量指标的定量模型。结果表明,酸价和极性组分的定标模型校正决定系数（R2）均为0.9974,校正标准差均方根（RMSEC）分别为0.111和0.359,预测标准差均方根（RMSEP）分别为0.171和0.562。盲样验证、精密度及准确度分析结果显示,酸价和极性组分的NIR预测值同真实值的相关方程的相关系数分别为0.9944、0.9761,应用所建模型预测同一煎炸油样品的酸价和极性组分的相对标准偏差（RSD）分别为0.934%和1.278%,表明所建模型对煎炸油样品的酸价和极性组分预测能力较好,且有较好的重现性。因此,基于近红外光谱定量模型,可以对西式快餐用煎炸油质量进行快速、准确地监测。     

三维荧光光谱法快速检测橄榄油中掺假廉价油

目的 建立三维荧光光谱结合机器学习快速检测橄榄油中掺假廉价油的方法。方法 采集橄榄油及掺入大豆油、玉米油、棕榈油三种不同浓度梯度油的荧光光谱数据,利用标准差标准化（standardscaler）、标准正态变换（standard normal variate,SNV）、归一化（normalize）三种光谱预处理方法,基于K近邻(K-nearest neighbor,KNN)、随机森林(random forest,RF)、支持向量机(support vector machine,SVM)、偏最小二乘法(partial least squares,PLS)和卷积神经网络(convolutional neural network,CNN) 5种机器学习方法,构建5种橄榄油定量掺假模型。结果 在定性模型中,基于PLS算法构建的模型效果最好,对3种掺假橄榄油的准确率为79%~97%,其中,在鉴定掺假大豆油的橄榄油中正确率高达97%。在构建的掺假油定量模型中,Standardscaler预处理结合RF算法,构建的定量模型最优,Rc2、Rp2、RMSEC、RMSEP最高,分别为1.00、0.99、0.01、0.02。结论 构建橄榄油掺假3种油的定性定量模型,并建立一种快速、实时、低成本的橄榄油掺假检测方法,能够准确判断是否掺入廉价油,并量化掺假程度,提供更全面的橄榄油质量评估。     

基于近红外光谱快速定量检测面粉中曲酸的方法建立

利用近红外光谱技术快速定量检测面粉中非法添加的褐变抑制剂曲酸。选取市场上常见3?种基本类型的面粉（高、中、低筋面粉）,分别制备曲酸质量分数为0.0%、0.5%、1.0%、3.0%、5.0%、10.0%的面粉样品,并采集其在1?000～2?400?nm波段下的光谱数据。对比不同预处理下高筋面粉样品数据所建偏最小二乘（partial least squares,PLS）回归模型效果,选取Savitzky-Golay一阶导数为最优预处理方法。采用区间偏最小二乘（interval partial least squares,iPLS）法选取1?088.8～1?153.5?nm为最佳光谱区间。结果表明,基于最佳光谱区间所建PLS回归模型预测效果优于基于全波段光谱数据所建模型。进一步,基于所选最优区间对中、低筋面粉和混合样品集分别建立PLS回归模型。高、中、低筋面粉及混合样品集基于最优区间的PLS模型的决定系数为0.949～0.972,标准误差为0.581%～0.830%,验证集标准偏差与预测标准偏差的比值为4.171～4.830。结果表明,基于最优区间的近红外光谱方法对不同类型面粉中曲酸质量分数为1.0%～10.0%的样品具有较好的预测结果,结合具有低检测限的化学检测方法,在对大批量样品的检测中可提高检测效率。     

基于高光谱成像技术的小白杏成熟度判别模型

为快速、准确检测小白杏的成熟度,该研究选择七成熟、八成熟、九成熟、十成熟的小白杏样本各120个,采用近红外高光谱成像系统采集样本的高光谱数据,进行去除噪声和剔除界外样本处理。然后使用均值中心化（mean centering,MC）、Savitzky-Golay卷积求导法（Savitzky-Golay derivative,S-G）、多元散射校正（multiplicative scatter correction,MSC）、标准正态变量变换（standard normal variate transformation,SNV）、归一化法5种方法分别对全波段和特征波段光谱进行预处理,采用光谱-理化值共生距离算法（sample set partitioning based on joint x-y distance,SPXY）、K-S法（Kennard-stone,K-S）、双向算法（Duplex）、交叉验证法、随机法将样本划分为校正集和验证集。最后用极限学习机（extreme learning machine,ELM）、支持向量机（support vector machine,SVM）、偏最小二乘法（partial least squares,PLS）、K最邻近法（K-nearest neighbor,KNN）、贝叶斯判别法建立不同的分类判别模型,比较各模型的识别率。结果表明,对小白杏成熟度定性判别模型,有以下最优组合：全波段+MSC+SPXY/Duplex/K-S/交叉验证/随机法+ELM/PLS/SVM/KNN、全波段+S-G/MSC/归一化/SNV+随机法+贝叶斯、全波段+S-G+SPXY/Duplex/K-S/交叉验证/随机法+ELM/PLS/SVM/KNN、全波段+归一化+SPXY/Duplex/K-S/交叉验证/随机法+PLS、特征波段+MSC+SPXY/Duplex/K-S/交叉验证/随机法+ELM/PLS/SVM/KNN/贝叶斯、特征波段+归一化+SPXY/Duplex/K-S/交叉验证/随机法+PLS。     

基于不同颜色模型定量预测永川秀芽在制品含水率

为定量预测永川秀芽在制品的含水率,基于不同颜色模型探究在制品的色泽变化,并结合偏最小二乘（partial least square,PLS）法建立含水率的定量预测模型。结果表明：在永川秀芽初制过程中,在制品的红绿度、蓝色通道均值增高,含水率和亮度、黄蓝度、红色通道均值、绿色通道均值、色调均值等15 个颜色模型分量降低,即色泽表现为变暗、变黄;通过热图与聚类分析,可将在制品分为2 个大类、4 个亚类,且理条工序对在制品含水率、色泽的影响最为显著;利用17 个颜色模型分量和PLS方法建立了含水率的定量预测模型,以校正集相关系数（Rc）、交互验证均方根误差（root-mean-square error of cross-validation,RMSECV）、预测集相关系数（Rp）、预测均方根误差（root-mean-square error of prediction,RMSEP）、相对分析误差（relative percent deviation,RPD）为评价指标。模型的Rc、Rp、RMSECV、RMSEP分别为0.979、0.980、0.044 7、0.044 3。RMSECV、RMSEP的差值仅为0.000 4,且RPD达到5.04,表明模型具有极好的预测能力和泛化能力,为实现永川秀芽在制品含水率的在线监测提供了一种新方法。     

基于高光谱成像技术和连续投影算法检测葡萄果皮花色苷含量

应用高光谱成像技术结合连续投影算法（SPA）实现葡萄果皮中花色苷含量的快速无损检测。采集60 组样本高光谱图像,获取样本光谱曲线,并采用多元散射校正预处理方法提高信噪比。然后采用SPA选择光谱变量,将其作为多元线性回归（MLR）、偏最小二乘（PLS）模型和BP神经网络（BPNN）的输入变量,分别建立SPAMLR、SPA-PLS和SPA-BPNN模型并与全光谱变量PLS模型相比较。结果表明,SPA-MLR、SPA-BPNN和SPA-PLS模型的预测精度均优于全光谱变量PLS模型,其中SPA-PLS模型获得了最佳预测结果,其预测相关系数Rp和预测均方根误差（RMSEP）分别为0.900 0和0.550 6。结果表明,利用近红外高光谱成像技术能够有效检测酿酒葡萄果皮中花色苷含量。     

气相色谱-质谱联用结合偏最小二乘法在生鲜牛、鸭混合肉含量判别分析中的应用

采用顶空固相微萃取（headspace solid-phase micro extraction,HS-SPME）和气相色谱-质谱联用（gas chromatography-mass spectrometry,GC-MS）技术提取和分析生鲜牛、鸭混合肉中的挥发性化合物,利用偏最小二乘（partial least squares,PLS）法对GC-MS数据进行建模,对不同比例生鲜牛、鸭混合肉进行判别分析。结果表明：生鲜牛、鸭混合肉中共鉴定出39 种挥发性化合物;PLS模型可以有效区分不同比例生鲜牛、鸭混合肉,利用训练集和验证集数据验证模型,评估均方差为3.455 9,预测均方差为4.068 9,PLS模型对不同比例生鲜牛、鸭混合肉的预测准确率较高。本方法在判别不同比例生鲜牛、鸭混合肉方面准确性较高。     

基于质谱与支持向量机的清香型白酒等级判别

不同等级白酒鉴别对控制白酒质量和保护消费者利益有重要意义,文中以牛栏山酒为例,研究清香型白酒质量等级鉴别方法。运用顶空固相微萃取质谱(HS-SPME-MS)技术获取三类不同等级的57个牛栏山酒样质荷比m/z55~191范围内的离子丰度值数据,分别进行偏最小二乘回归分析(PLS)和主成分回归分析(PCR),其中PLS模型的预测结果明显优于PCR。同时PLS与PCR模型的回归系数用于选择重要特征离子,其中PLS与PCR回归系数法分别选择了12和10个离子,用选择的离子变量构建支持向量机(SVM)模型,模型对测试集的预测准确率分别为80%和86.7%,其中PCR回归系数法选择的特征离子为m/z 71、103、104、106、127、149、161、179、183和184。     

基于集成学习模型的交通事故严重程度时空预测

为探究区域交通事故时空特征,精准预测事故严重程度,给交通运输主管部门提供决策支持,以英国交通事故统计数据作为研究基础,首先,将交通事故时空特征数据转化为网格化数据,并对空间特征进行二维卷积,利用时间特征合并二维卷积为三维卷积,解决网格冲突问题;其次,利用卷积神经网络（CNN）、长短期记忆人工神经网络（LSTM）模型的工作原理建立Stacking模型的基学习器和元学习器;最后,输出结果传入分类与回归树（CART）,构建完整的事故严重程度预测集成学习模型。研究结果表明,集成学习模型较单一模型对预测效果更优,其AUC比CNN、LSTM和Conv-LSTM单一模型预测分别提升0.02、0.04和0.01;最终决策树选择中,CART决策树比随机森林（RF）和梯度提升决策树（GBDT）预测效果更优;预测结果在时间纬度上,“严重事故”事件占比较实际低3.95%,在空间纬度上,预测热力区域在0.5～1区间范围内与实际接近。     

Reliable identification of grapevine rootstock varieties using RAPD PCR on woody samples

Since grapevine ( Vitis spp .) rootstock material is being traded increasingly as disbudded woody material a lack of distinctive morphological features on such material necessitates an alternative and reliable means of identification. Methods described here were developed for rapid and efficient extraction of DNA from woody samples rich in phenolic compounds and polysaccharides, and for subsequent identification of varieties by RAPD PCR. Using these methods, and with the application of only one selected RAPD primer, we were able to differentiate sixteen rootstock varieties, including the seven varieties most commonly used in Germany. Problems commonly encountered with reproducibility of RAPD patterns were avoided by choosing primers with a dinucleotide sequence and a high G/C content that allowed a rather high annealing temperature of 45°C. Methods described here should also be useful for other horticultural crops, especially those with woody tissues rich in phenolic compounds and polysaccharides.     

An internet compendium of analytical methods and spectroscopic information for monomers and additives used in food packaging plastics

An internet website (http://cpf.jrc.it/smt/) has been produced as a means of dissemination of methods of analysis and supporting spectroscopic information on monomers and additives used for food contact materials (principally packaging). The site which is aimed primarily at assisting food control laboratories in the European Union contains analytical information on monomers, starting substances and additives used in the manufacture of plastics materials. A searchable index is provided giving PM and CAS numbers for each of 255 substances. For each substance a data sheet gives regulatory information, chemical structures, physico-chemical information and background information on the use of the substance in particular plastics, and the food packaging applications. For monomers and starting substances (155 compounds) the infra-red and mass spectra are provided, and for additives (100 compounds); additionally proton NMR are available for about 50% of the entries. Where analytical methods have been developed for determining these substances as residual amounts in plastics or as trace amounts in food simulants these methods are also on the website. All information is provided in portable document file (PDF) format which means that high quality copies can be readily printed, using freely available Adobe Acrobat Reader software. The website will in future be maintained and up-dated by the European Commission's Joint Research Centre (JRC) as new substances are authorized for use by the European Commission (DG-ENTR formerly DGIII). Where analytical laboratories (food control or other) require reference substances these can be obtained free-ofcharge from a reference collection housed at the JRC and maintained in conjunction with this website compendium.     

Aroma characterization of various apricot varieties using headspace–solid phase microextraction combined with gas chromatography–mass spectrometry and gas chromatography–olfactometry

The characterization of the aromatic profile of several apricot cultivars with molecular tracers in order to obtain objective data concerning the aromatic quality of this fruit was undertaken using headspace–solid phase microextraction (HS–SPME). Six apricot cultivars were selected according to their organoleptic characteristics: Iranien, Orangered, Goldrich, Hargrand, Rouge du Roussillon and A4025. The aromatic intensity of these varieties measured by HS–SPME–Olfactometry were defined and classified according to the presence and the intensity of grassy, fruity and apricot like notes. In the six varieties, 23 common volatile compounds were identified by HS–SPME–GC–MS. Finally, 10 compounds, ethyl acetate, hexyl acetate, limonene, β-cyclocitral, γ-decalactone, 6-methyl-5-hepten-2-one, linalool, β-ionone, menthone and (E)-hexen-2-al were recognized by HS–SPME–GC–O as responsible of the aromatic notes involved in apricot aroma and considered as molecular tracers of apricot aromatic quality which could be utilized to discriminate apricot varieties.     

Tests of potential functional barriers for laminated multilayer food packages. Part I: Low molecular weight permeants

The advent of the functional barrier concept in food packaging has brought with it a requirement for fast tests of permeation through potential barrier materials. In such tests it would be convenient for both foodstuffs and materials below the functional barrier (sub-barrier materials) to be represented by standard simulants. By means of inverse gas chromatography, liquid paraffin spiked with appropriate permeants was considered as a potential simulant of sub-barrier materials based on polypropylene (PP) or similar polyolefins. Experiments were performed to characterize the kinetics of the permeation of low molecular weight model permeants (octene, toluene and isopropanol) from liquid paraffin, through a surrogate potential functional barrier (25 μm-thick oriented PP) into the food simulants olive oil and 3% (w/v) acetic acid. These permeation results were interpreted in terms of three permeation kinetic models regarding the solubility of a particular model permeant in the post-barrier medium (i.e. the food simulant). The results obtained justify the development and evaluation of liquid sub-barrier simulants that would allow flexible yet rigorous testing of new laminated multilayer packaging materials.     

Textural Properties of Cold-set Gels Induced from Heat-denatured Whey Protein Isolates

A 9% whey protein (WP) isolate solution at pH 7.0 was heat-denatured at 80°C for 30 min. Size-exclusion HPLC showed that native WP formed soluble aggregates after heat-treatment. Additions of CaCl2 (10–40 mM), NaCl (50–400 mM) or glucono-delta-lactone (GDL, 0.4–2.0%, w/v) or hydrolysis by a protease from Bacillus licheniformis caused gelation of the denatured solution at 45°C. Textural parameters, hardness, adhesiveness, and cohesiveness of the gels so formed changed markedly with concentration of added salts or pH by added GDL. Maximum gel hardness occurred at 200 mM NaCl or pH 4.7. Increasing CaCl2 concentration continuously increased gel hardness. Generally, GDL-induced gels were harder than salt-induced gels, and much harder than the protease-induced gel.     

Occurrence of bisphenol-F-diglycidyl ether (BFDGE) in fish canned in oil

The levels of bisphenol-F-diglycidyl ether (BFDGE) were quantified as part of a European survey on the migration of residues of epoxy resins into oil from canned fish. The contents of BFDGE in cans, lids and fish collected from all 15 Member States of the European Union and Switzerland were analysed in 382 samples. Cans and lids were separately extracted with acetonitrile. The extraction from fish was carried out with hexane followed by re-extraction with acetonitrile. The analysis was performed by reverse phase HPL C with fluorescence detection. BFDGE could be detected in 12% of the fish, 24% of the cans and 18% of the lids. Only 3% of the fish contained BFDGE in concentrations considerably above 1mg/kg. In addition to the presented data, a comparison was made with the levels of BADGE (bisphenol-A-diglycidyl ether)analysed in the same products in the context of a previous study.     

Analysis of benzo[a]pyrene in spiked fatty foods by second derivative synchronous spectrofluorimetry after microwave-assisted treatment of samples

A simple, rapid and inexpensive method has been developed for the determination of benzo[a     

The development of reference materials for paralytic shellfish poisoning toxins in lyophilized mussel. II: Certification study

This paper describes the second part of a project undertaken to develop certified mussel reference materials for paralytic shellfish poisoning toxins. In the first part two interlaboratory studies were undertaken to investigate the performance of the analytical methodology for several PSP toxins, in particular saxitoxin and decarbamoyl-saxitoxin in lyophilized mussels, and to set criteria for the acceptance of results to be applied during the certification exercise. Fifteen laboratories participated in this certification study and were asked to measure saxitoxin and decarbamoyl-saxitoxin in rehydrated lyophilized mussel material and in a saxitoxin-enriched mussel material. The participants were allowed to use a method of their choice but with an extraction procedure to be strictly followed. The study included extra experiments to verify the detection limits for both saxitoxin and decarbamoyl-saxitoxin. Most participants (13 of 15) were able to meet all the criteria set for the certification study. Results for saxitoxin.2HCl yielded a certified mass fraction of <0.07 mg/kg in the rehydrated lyophilized mussels. Results obtained for decarbamoyl-saxitoxin.2HCl yielded a certified mass fraction of 1.59+/-0.20 mg/kg. The results for saxitoxin.2HCl in enriched blank mussel yielded a certified mass fraction of 0.48 +/- 0.06 mg/kg. These certified reference materials for paralytic shellfish poisoning toxins in lyophilized mussel material are the first available for laboratories to test their method for accuracy and performance.     

Contribution of European research to risk analysis

The European Commission's, Quality of Life Research Programme, Key Action 1—Health, Food & Nutrition is mission-oriented and aims, amongst other things, at providing a healthy, safe and high-quality food supply leading to reinforced consumer confidence in the safety of European food. Its objectives also include the enhancing of the competitiveness of the European food supply. Key Action 1 is currently supporting a number of different types of European collaborative projects in the area of risk analysis. The objectives of these projects range from the development and validation of prevention strategies including the reduction of consumers risks; development and validation of new modelling approaches; harmonization of risk assessment principles, methodologies, and terminology; standardization of methods and systems used for the safety evaluation of transgenic food; providing of tools for the evaluation of human viral contamination of shellfish and quality control; new methodologies for assessing the potential of unintended effects of genetically modified (genetically modified) foods; development of a risk assessment model for Cryptosporidium parvum related to the food and water industries; to the development of a communication platform for genetically modified organism, producers, retailers, regulatory authorities and consumer groups to improve safety assessment procedures, risk management strategies and risk communication; development and validation of new methods for safety testing of transgenic food; evaluation of the safety and efficacy of iron supplementation in pregnant women; evaluation of the potential cancer-preventing activity of pro- and pre-biotic ('synbiotic') combinations in human volunteers. An overview of these projects is presented here.