On the differential benchmarking of promotional efficiency with machine learning modelling (II): Practical applications

The assessment of promotional sales with models constructed by machine learning techniques is arousing interest due, among other reasons, to the current economic situation leading to a more complex environment of simultaneous and concurrent promotional activities. An operative model diagnosis procedure was previously proposed in the companion paper, which can be readily used both for agile decision making on the architecture and implementation details of the machine learning algorithms, and for differential benchmarking among models. In this paper, a detailed example of model analysis is presented for two representative databases with different promotional behaviour, namely, a non-seasonal category (milk) and a heavily seasonal category (beer). The performance of four well-known machine learning techniques with increasing complexity is analyzed in detail here. In particular, k-Nearest Neighbours, General Regression Neural Networks, Multilayer Perceptron (MLP), and Support Vector Machines (SVM), are differentially compared. Present paper evaluates these techniques along the experiments described for both categories when applying the methodological findings obtained in the companion paper. We conclude that some elements included in the architecture are not essential for a good performance of the machine learning promotional models, such as the semiparametric nature of the kernel in SVM models, whereas other can be strongly dependent of the database, such as the convenience of multiple output models in MLP regression schemes. Additionally, the specificity of the behaviour of certain categories and product ranges determines the need to establish suitable and specific procedures for a better prediction and feature extraction.     

Learning Photometric Invariance for Object Detection

Color is a powerful visual cue in many computer vision applications such as image segmentation and object recognition. However, most of the existing color models depend on the imaging conditions that negatively affect the performance of the task at hand. Often, a reflection model (e.g., Lambertian or dichromatic reflectance) is used to derive color invariant models. However, this approach may be too restricted to model real-world scenes in which different reflectance mechanisms can hold simultaneously.     

Mobile crowdsensing with mobile agents

We introduce mobile agents for mobile crowdsensing. Crowdsensing campaigns are designed through different roles that are implemented as mobile agents. The role-based tasks of mobile agents include collecting data, analyzing data and sharing data in the campaign. Mobile agents execute and control the campaign autonomously as a multi-agent system and migrate in the opportunistic network of participants’ devices. Mobile agents take into account the available resources in the devices and match participants’ privacy requirements to the campaign requirements. Sharing of task results in real-time facilitates cooperation towards the campaign goal while maintaining a selected global measure, such as energy efficiency. We discuss current challenges in crowdsensing and propose mobile agent based solutions for campaign execution and monitoring, addressing data collection and participant-related issues. We present a software framework for mobile agents-based crowdsensing that is seamlessly integrated into the Web. A set of simulations are conducted to compare mobile agent-based campaigns with existing crowdsensing approaches. We implemented and evaluated a small-scale real-world mobile agent based campaign for pedestrian flock detection. The simulation and evaluation results show that mobile agent based campaigns produce comparable results with less energy consumption when the number of agents is relatively small and enables in-network data processing with sharing of data and task results with insignificant overhead.     

A multiobjective approach based on artificial bee colony for the static routing and wavelength assignment problem

Nowadays, the most promising technology for designing optical networks is the wavelength division multiplexing. This technique divides the huge bandwidth of an optical fiber link into different wavelengths, providing different available channels per link. However, a problem comes up when it is necessary to interconnect a set of traffic demands. This problem is known as routing and wavelength assignment problem, and due to its complexity (NP-hard problem) it is very suitable for being solved using evolutionary computation. The selected heuristics is the artificial bee colony algorithm, an heuristics based on the behavior of honeybee foraging for nectar. Therefore, we have applied multiobjective optimization to solve the static routing and wavelength assignment problem, and adapted this algorithm to the multiobjective context. New results have been obtained that significantly improve those published in previous researches.     

A new methodology to improve interpretability in neuro-fuzzy TSK models

The present paper puts forward a methodology which allows increasing interpretability of TSK models identified by means of neuro-fuzzy techniques, although it shall also be applicable to models identified through other hybrid or different techniques. With this purpose, this paper puts forward a method which allows oriented adjustment of the rules’ precedent and consequent parameters in TSK models. The methodology extends the adaptive phase with an adjustment phase (or fine tuning phase) based on overlap ratio and overlap area, where the gradient descendent algorithm is used to adjust precisely the adapted parameters in the fuzzy model. The adjustment based on the overlap ratio is applied to the parameters defining the rules’ precedent and consequent parts. The overlap area becomes a more precise tuning of parameters of precedent part of rules. After the adaptation of the neuro-fuzzy model by means of the developed methodology, the model acquires a clear physical meaning enabling its immediate linguistic interpretation. Finally, some examples are given to prove the validity of the developed methodology.     

A multivariable control scheme in a two-stage anaerobic digestion system described by partial differential equations

This paper deals with the design and application of a nonlinear multivariable controller in an anaerobic digestion system (AD) carried out in two interconnected fixed bed bioreactors. The proposed control scheme is derived from a mathematical model of the AD system described by a set of partial differential equations and consists of an estimator and two nonlinear control laws. The first law is developed to regulate the volatile fatty acids in the first bioreactor while the second aims at maintaining the chemical oxygen demand at predetermined set-points in the second bioreactor. The performance of the control algorithm is evaluated via numerical simulations in the face of load disturbances, parameter kinetic uncertainties and set-point changes. Stability and convergence properties of the proposed control scheme are also addressed in this paper.     

Improving spectral results in a GTS context

This paper shows the advantages of post-processing spectral classifications in a Geographical Information System (GIS) context in order to improve results. A maximum-likelihood algorithm was used to classify(both supervised and non-supervised) a Landsat TM sub-image in Central Mexico. Purely spectral processing yielded poor accuracy results, showing the spectral limitation to distinguish classes; as a consequence, merging classes was necessary in order to increase accuracy (from less than 55 to 82 per cent). GIS rules were finally applied based on ancillary data (terrain mapping units and elevation data) improving the final accuracy to 88.2 and 83.0 per cent (supervised and non-supervised classifications).     

A parallel cooperative team of multiobjective evolutionary algorithms for motif discovery

When solving a wide range of complex scenarios of a given optimization problem, it is very difficult, if not impossible, to develop a single technique or algorithm that is able to solve all of them adequately. In this case, it is necessary to combine several algorithms by applying the most appropriate one in each case. Parallel computing can be used to improve the quality of the solutions obtained in a cooperative algorithms model. Exchanging information between parallel cooperative algorithms will alter their behavior in terms of solution searching, and it may be more effective than a sequential metaheuristic. For demonstrating this, a parallel cooperative team of four multiobjective evolutionary algorithms based on OpenMP is proposed for solving different scenarios of the Motif Discovery Problem (MDP), which is an important real-world problem in the biological domain. As we will see, the results show that the application of a properly configured parallel cooperative team achieves high quality solutions when solving the addressed problem, improving those achieved by the algorithms executed independently for a much longer time.     

A multi-start tabu search method for a single-machine scheduling problem with periodic maintenance and sequence-dependent set-up times

In this paper we study a problem of sequencing jobs in a machine with programmed preventive maintenance and sequence-dependent set-up times. The problem combines two NP-hard problems, so we propose a heuristic method for solving it, which hybridizes multi-start strategies with Tabu Search. We compare our method with the only published metaheuristic algorithm for this problem on a set of 420 instances. The comparison favors the method developed in this work, showing that is able to find high quality solutions in very short computational times.