Fast adaptive real-time classification for data streams with concept drift

An important application of Big Data Analytics is the real-time analysis of streaming data. Streaming data imposes unique challenges to data mining algorithms, such as concept drifts, the need to analyse the data on the fly due to unbounded data streams and scalable algorithms due to potentially high throughput of data. Real-time classification algorithms that are adaptive to concept drifts and fast exist, however, most approaches are not naturally parallel and are thus limited in their scalability. This paper presents work on the Micro-Cluster Nearest Neighbour (MC-NN) classifier. MC-NN is based on an adaptive statistical data summary based on Micro-Clusters. MC-NN is very fast and adaptive to concept drift whilst maintaining the parallel properties of the base KNN classifier. Also MC-NN is competitive compared with existing data stream classifiers in terms of accuracy and speed.

Heathland restoration on former agricultural land: effects of artificial acidification on the availability and uptake of toxic metal cations

Lowland heath is an internationally important habitat type that has greatly declined in abundance throughout Western Europe. In recent years this has led to a growing interest in the restoration of heathland on agricultural land. This generally requires the use of chemical treatments to return soil chemical conditions to those appropriate for the support of heathland ecosystems. However, the potential for negative impacts on the environment due to the potential of these treatments to increase the availability of trace metals via raised soil acidity requires investigation. A large-scale field study investigated the effect of two chemical treatments used in heathland restoration, elemental sulphur and ferrous sulphate, on soil acidity and whether it is possible to predict the effect of the treatments on availability of two potentially toxic cations (Al and Cd) in the soil along with their subsequent accumulation in the shoots of the grass Agrostis capillaris. Results showed that both treatments decreased soil pH, but that only elemental sulphur produced a pH similar to heathland soil. The availability of Al, measured by extraction with 1 M ammonium nitrate, could not be predicted by soil pH, depth in the soil and total Al concentration in the soil. By contrast, availability of Cd could be predicted from these three variables. Concentrations of both Al and Cd in the shoots of A. capillaris showed no significant relationship with the extractable concentration in the soil. Results are discussed in light of the possible environmental impacts of the chemical restoration techniques.

Valorisation of side streams from wheat milling and confectionery industries for consolidated production and extraction of microbial lipids

Crude enzymes produced via solid state fermentation (SSF) using wheat milling by-products have been employed for both fermentation media production using flour-rich waste (FRW) streams and lysis of Rhodosporidium toruloides yeast cells. Filter sterilization of crude hydrolysates was more beneficial than heat sterilization regarding yeast growth and microbial oil production. The initial carbon to free amino nitrogen ratio of crude hydrolysates was optimized (80.2 g/g) in fed-batch cultures of R. toruloides leading to a total dry weight of 61.2 g/L with microbial oil content of 61.8 % (w/w). Employing a feeding strategy where the glucose concentration was maintained in the range of 12.2 – 17.6 g/L led to the highest productivity (0.32 g/L∙h). The crude enzymes produced by SSF were utilised for yeast cell treatment leading to simultaneous release of around 80% of total lipids in the broth and production of a hydrolysate suitable as yeast extract replacement.

Rapid, automated detection of stem canker symptoms in woody perennials using artificial neural network analysis

Background Pseudomonas syringae can cause stem necrosis and canker in a wide range of woody species including cherry, plum, peach, horse chestnut and ash. The detection and quantification of lesion progression over time in woody tissues is a key trait for breeders to select upon for resistance. Results In this study a general, rapid and reliable approach to lesion quantification using image recognition and an artificial neural network model was developed. This was applied to screen both the virulence of a range of P. syringae pathovars and the resistance of a set of cherry and plum accessions to bacterial canker. The method developed was more objective than scoring by eye and allowed the detection of putatively resistant plant material for further study. Conclusions Automated image analysis will facilitate rapid screening of material for resistance to bacterial and other phytopathogens, allowing more efficient selection and quantification of resistance responses.

Formulation of fermentation media from flour-rich waste streams for microbial lipid production by Lipomyces starkeyi

Flour-rich waste (FRW) and by-product streams generated by bakery, confectionery and wheat milling plants could be employed as the sole raw materials for generic fermentation media production, suitable for microbial oil synthesis. Wheat milling by-products were used in solid state fermentations (SSF) of Aspergillus awamori for the production of crude enzymes, mainly glucoamylase and protease. Enzyme-rich SSF solids were subsequently employed for hydrolysis of FRW streams into nutrient-rich fermentation media. Batch hydrolytic experiments using FRW concentrations up to 205 g/L resulted in higher than 90%(w/w) starch to glucose conversion yields and 40% (w/w) total Kjeldahl nitrogen to free amino nitro-gen conversion yields. Starch to glucose conversion yields of 98.2, 86.1 and 73.4% (w/w) were achieved when initial FRW concentrations of 235, 300 and 350 g/L were employed in fed-batch hydrolytic experiments, respectively. Crude hydrolysates were used as fermentation media in shake flask cultures with the oleaginous yeast Lipomyces starkeyi DSM 70296 reaching a total dry weight of 30.5 g/L with a microbial oil content of 40.4% (w/w), higher than that achieved in synthetic media. Fed-batch bioreactor cultures led to a total dry weight of 109.8 g/L with a microbial oil content of 57.8% (w/w) and productivity of 0.4 g/L/h.