Abstract: Health and Social care (HSc) services planning and scheduling are facing unprecedented challenges, due to the pandemic pressure and also suffer from unplanned spending that is negatively impacted by the global financial crisis. Data-driven approaches can help to improve policies, plan and design services provision schedules using algorithms that assist healthcare managers to face unexpected demands using fewer resources. The paper discusses services packing using statistical significance tests and machine learning (ML) to evaluate demands similarity and coupling. This is achieved by predicting the range of the demand (class) using ML methods such as Classification and Regression Trees (CART), Random Forests (RF), and Logistic Regression (LGR). The significance tests Chi-Squared and Student’s test are used on data over a 39 years span for which data exist for services delivered in Scotland. The demands are associated using probabilities and are parts of statistical hypotheses. These hypotheses, as their NULL part, assume that the target demand is statistically dependent on other services’ demands. This linking is checked using the data. In addition, ML methods are used to linearly predict the above target demands from the statistically found associations and extend the linear dependence of the target’s demand to independent demands forming, thus, groups of services. Statistical tests confirmed ML coupling and made the prediction statistically meaningful and proved that a target service can be matched reliably to other services while ML showed that such marked relationships can also be linear ones. Zero padding was used for missing years records and illustrated better such relationships both for limited years and for the entire span offering long-term data visualizations while limited years periods explained how well patients numbers can be related in short periods of time or that they can change over time as opposed to behaviours across more years. The prediction performance of the associations were measured using metrics such as Receiver Operating Characteristic (ROC), Area Under Curve (AUC) and Accuracy (ACC) as well as the statistical tests Chi-Squared and Student. Co-plots and comparison tables for the RF, CART, and LGR methods as well as the p-value from tests and Information Exchange (IE/MIE) measures are provided showing the relative performance of ML methods and of the statistical tests as well as the behaviour using different learning ratios. The impact of k-neighbours classification (k-NN), Cross-Correlation (CC) and C-Means (CM) first groupings was also studied over limited years and for the entire span. It was found that CART was generally behind RF and LGR but in some interesting cases, LGR reached an AUC = 0 falling below CART, while the ACC was as high as 0.912 showing that ML methods can be confused by zero-padding or by data’s irregularities or by the outliers. On average, 3 linear predictors were sufficient, LGR was found competing well RF and CART followed with the same performance at higher learning ratios. Services were packed only when a significance level (p-value) of their association coefficient was more than 0.05. Social factors relationships were observed between home care services and treatment of old people, low birth weights, alcoholism, drug abuse, and emergency admissions. The work found that different HSc services can be well packed as plans of limited duration, across various services sectors, learning configurations, as confirmed by using statistical hypotheses.
Abstract: Early detection of anomalies in data centers is important to reduce downtimes and the costs of periodic maintenance. However, there is little research on this topic and even fewer on the fusion of sensor data for the detection of abnormal events. The goal of this paper is to propose a method for anomaly detection in data centers by combining sensor data (temperature, humidity, power) and deep learning models. The model described in the paper uses one autoencoder per sensor to reconstruct the inputs. The auto-encoders contain Long-Short Term Memory (LSTM) layers and are trained using the normal samples of the relevant sensors selected by correlation analysis. The difference signal between the input and its reconstruction is then used to classify the samples using feature extraction and a random forest classifier. The data measured by the sensors of a data center between January 2019 and May 2020 are used to train the model, while the data between June 2020 and May 2021 are used to assess it. Performances of the model are assessed a posteriori through F1-score by comparing detected anomalies with the data center’s history. The proposed model outperforms the state-of-the-art reconstruction method, which uses only one autoencoder taking multivariate sequences and detects an anomaly with a threshold on the reconstruction error, with an F1-score of 83.60% compared to 24.16%.
Abstract: Promotion is a key element in the retail business. Thus, analysis of promotions to quantify their effectiveness in terms of Revenue and/or Margin is an essential activity in the retail industry. However, measuring the sales/revenue uplift is based on estimations, as the actual sales/revenue without the promotion is not present. Further, the presence of Halo and Cannibalization in a multiple parallel promotions’ scenario complicates the problem. Calculating Baseline by considering inter-brand/competitor items or using Halo and Cannibalization's impact on Revenue calculations by considering Baseline as an interpretation of items’ unit sales in neighboring nonpromotional weeks individually may not capture the overall Revenue uplift in the case of multiple parallel promotions. Hence, this paper proposes a Machine Learning based method for calculating the Revenue uplift by considering the Halo and Cannibalization impact on the Baseline and the Revenue. In the first section of the proposed methodology, Baseline of an item is calculated by incorporating the impact of the promotions on its related items. In the later section, the Revenue of an item is calculated by considering both Halo and Cannibalization impacts. Hence, this methodology enables correct calculation of the overall Revenue uplift due a given promotion.
Abstract: In this study, a multivariate analysis of potato spectroscopic data was presented to detect the presence of brown rot disease or not. Near-Infrared (NIR) spectroscopy (1,350-2,500 nm) combined with multivariate analysis was used as a rapid, non-destructive technique for the detection of brown rot disease in potatoes. Spectral measurements were performed in 565 samples, which were chosen randomly at the infection place in the potato slice. In this study, 254 infected and 311 uninfected (brown rot-free) samples were analyzed using different advanced statistical analysis techniques. The discrimination performance of different multivariate analysis techniques, including classification, pre-processing, and dimension reduction, were compared. Applying a random forest algorithm classifier with different pre-processing techniques to raw spectra had the best performance as the total classification accuracy of 98.7% was achieved in discriminating infected potatoes from control.
Abstract: Engagement is one of the most important factors in determining successful outcomes and deep learning in students. Existing approaches to detect student engagement involve periodic human observations that are subject to inter-rater reliability. Our solution uses real-time multimodal multisensor data labeled by objective performance outcomes to infer the engagement of students. The study involves four students with a combined diagnosis of cerebral palsy and a learning disability who took part in a 3-month trial over 59 sessions. Multimodal multisensor data were collected while they participated in a continuous performance test. Eye gaze, electroencephalogram, body pose, and interaction data were used to create a model of student engagement through objective labeling from the continuous performance test outcomes. In order to achieve this, a type of continuous performance test is introduced, the Seek-X type. Nine features were extracted including high-level handpicked compound features. Using leave-one-out cross-validation, a series of different machine learning approaches were evaluated. Overall, the random forest classification approach achieved the best classification results. Using random forest, 93.3% classification for engagement and 42.9% accuracy for disengagement were achieved. We compared these results to outcomes from different models: AdaBoost, decision tree, k-Nearest Neighbor, naïve Bayes, neural network, and support vector machine. We showed that using a multisensor approach achieved higher accuracy than using features from any reduced set of sensors. We found that using high-level handpicked features can improve the classification accuracy in every sensor mode. Our approach is robust to both sensor fallout and occlusions. The single most important sensor feature to the classification of engagement and distraction was shown to be eye gaze. It has been shown that we can accurately predict the level of engagement of students with learning disabilities in a real-time approach that is not subject to inter-rater reliability, human observation or reliant on a single mode of sensor input. This will help teachers design interventions for a heterogeneous group of students, where teachers cannot possibly attend to each of their individual needs. Our approach can be used to identify those with the greatest learning challenges so that all students are supported to reach their full potential.
Abstract: This paper presents a Machine Learning (ML) approach to support Meningitis diagnosis in patients at a children’s hospital in Sao Paulo, Brazil. The aim is to use ML techniques to reduce the use of invasive procedures, such as cerebrospinal fluid (CSF) collection, as much as possible. In this study, we focus on predicting the probability of Meningitis given the results of a blood and urine laboratory tests, together with the analysis of pain or other complaints from the patient. We tested a number of different ML algorithms, including: Adaptative Boosting (AdaBoost), Decision Tree, Gradient Boosting, K-Nearest Neighbors (KNN), Logistic Regression, Random Forest and Support Vector Machines (SVM). Decision Tree algorithm performed best, with 94.56% and 96.18% accuracy for training and testing data, respectively. These results represent a significant aid to doctors in diagnosing Meningitis as early as possible and in preventing expensive and painful procedures on some children.
Abstract: Unidentified Flying Objects (UFOs) have been an interesting topic for most enthusiasts and hence people all over the United States report such findings online at the National UFO Report Center (NUFORC). Some of these reports are a hoax and among those that seem legitimate, our task is not to establish that these events confirm that they indeed are events related to flying objects from aliens in outer space. Rather, we intend to identify if the report was a hoax as was identified by the UFO database team with their existing curation criterion. However, the database provides a wealth of information that can be exploited to provide various analyses and insights such as social reporting, identifying real-time spatial events and much more. We perform analysis to localize these time-series geospatial events and correlate with known real-time events. This paper does not confirm any legitimacy of alien activity, but rather attempts to gather information from likely legitimate reports of UFOs by studying the online reports. These events happen in geospatial clusters and also are time-based. We look at cluster density and data visualization to search the space of various cluster realizations to decide best probable clusters that provide us information about the proximity of such activity. A random forest classifier is also presented that is used to identify true events and hoax events, using the best possible features available such as region, week, time-period and duration. Lastly, we show the performance of the scheme on various days and correlate with real-time events where one of the UFO reports strongly correlates to a missile test conducted in the United States.
Abstract: A major challenge in medical studies, especially those that are longitudinal, is the problem of missing measurements which hinders the effective application of many machine learning algorithms. Furthermore, recent Alzheimer's Disease studies have focused on the delineation of Early Mild Cognitive Impairment (EMCI) and Late Mild Cognitive Impairment (LMCI) from cognitively normal controls (CN) which is essential for developing effective and early treatment methods. To address the aforementioned challenges, this paper explores the potential of using the eXtreme Gradient Boosting (XGBoost) algorithm in handling missing values in multiclass classification. We seek a generalized classification scheme where all prodromal stages of the disease are considered simultaneously in the classification and decision-making processes. Given the large number of subjects (1631) included in this study and in the presence of almost 28% missing values, we investigated the performance of XGBoost on the classification of the four classes of AD, NC, EMCI, and LMCI. Using 10-fold cross validation technique, XGBoost is shown to outperform other state-of-the-art classification algorithms by 3% in terms of accuracy and F-score. Our model achieved an accuracy of 80.52%, a precision of 80.62% and recall of 80.51%, supporting the more natural and promising multiclass classification.
Abstract: Motion recognition from videos is actually a very
complex task due to the high variability of motions. This paper
describes the challenges of human motion recognition, especially
motion representation step with relevant features. Our descriptor
vector is inspired from Laban Movement Analysis method. We
propose discriminative features using the Random Forest algorithm
in order to remove redundant features and make learning algorithms
operate faster and more effectively. We validate our method on
MSRC-12 and UTKinect datasets.
Abstract: Crop yield prediction is a paramount issue in
agriculture. The main idea of this paper is to find out efficient
way to predict the yield of corn based meteorological records.
The prediction models used in this paper can be classified into
model-driven approaches and data-driven approaches, according to
the different modeling methodologies. The model-driven approaches are based on crop mechanistic
modeling. They describe crop growth in interaction with their
environment as dynamical systems. But the calibration process of
the dynamic system comes up with much difficulty, because it
turns out to be a multidimensional non-convex optimization problem.
An original contribution of this paper is to propose a statistical
methodology, Multi-Scenarios Parameters Estimation (MSPE), for the
parametrization of potentially complex mechanistic models from a
new type of datasets (climatic data, final yield in many situations).
It is tested with CORNFLO, a crop model for maize growth. On the other hand, the data-driven approach for yield prediction
is free of the complex biophysical process. But it has some strict
requirements about the dataset.
A second contribution of the paper is the comparison of these
model-driven methods with classical data-driven methods. For this
purpose, we consider two classes of regression methods, methods
derived from linear regression (Ridge and Lasso Regression, Principal
Components Regression or Partial Least Squares Regression) and
machine learning methods (Random Forest, k-Nearest Neighbor,
Artificial Neural Network and SVM regression).
The dataset consists of 720 records of corn yield at county scale
provided by the United States Department of Agriculture (USDA) and
the associated climatic data. A 5-folds cross-validation process and
two accuracy metrics: root mean square error of prediction(RMSEP),
mean absolute error of prediction(MAEP) were used to evaluate the
crop prediction capacity.
The results show that among the data-driven approaches, Random
Forest is the most robust and generally achieves the best prediction
error (MAEP 4.27%). It also outperforms our model-driven approach
(MAEP 6.11%). However, the method to calibrate the mechanistic
model from dataset easy to access offers several side-perspectives.
The mechanistic model can potentially help to underline the stresses
suffered by the crop or to identify the biological parameters of interest
for breeding purposes. For this reason, an interesting perspective is
to combine these two types of approaches.
Abstract: Earthquake is an inevitable catastrophic natural disaster. The damages of buildings and man-made structures, where most of the human activities occur are the major cause of casualties from earthquakes. A comparison of optical and SAR data is presented in the case of Kathmandu valley which was hardly shaken by 2015-Nepal Earthquake. Though many existing researchers have conducted optical data based estimated or suggested combined use of optical and SAR data for improved accuracy, however finding cloud-free optical images when urgently needed are not assured. Therefore, this research is specializd in developing SAR based technique with the target of rapid and accurate geospatial reporting. Should considers that limited time available in post-disaster situation offering quick computation exclusively based on two pairs of pre-seismic and co-seismic single look complex (SLC) images. The InSAR coherence pre-seismic, co-seismic and post-seismic was used to detect the change in damaged area. In addition, the ground truth data from field applied to optical data by random forest classification for detection of damaged area. The ground truth data collected in the field were used to assess the accuracy of supervised classification approach. Though a higher accuracy obtained from the optical data then integration by optical-SAR data. Limitation of cloud-free images when urgently needed for earthquak evevent are and is not assured, thus further research on improving the SAR based damage detection is suggested. Availability of very accurate damage information is expected for channelling the rescue and emergency operations. It is expected that the quick reporting of the post-disaster damage situation quantified by the rapid earthquake assessment should assist in channeling the rescue and emergency operations, and in informing the public about the scale of damage.
Abstract: Predictive data analysis and modeling involving machine learning techniques become challenging in presence of too many explanatory variables or features. Presence of too many features in machine learning is known to not only cause algorithms to slow down, but they can also lead to decrease in model prediction accuracy. This study involves housing dataset with 79 quantitative and qualitative features that describe various aspects people consider while buying a new house. Boruta algorithm that supports feature selection using a wrapper approach build around random forest is used in this study. This feature selection process leads to 49 confirmed features which are then used for developing predictive random forest models. The study also explores five different data partitioning ratios and their impact on model accuracy are captured using coefficient of determination (r-square) and root mean square error (rsme).
Abstract: Complementary and Alternative Medicine (CAM) techniques are quite popular and effective for chronic diseases. Iridology is more than 150 years old CAM technique which analyzes the patterns, tissue weakness, color, shape, structure, etc. for disease diagnosis. The objective of this paper is to validate the use of iridology for the diagnosis of the diabetes. The suggested model was applied in a systemic disease with ocular effects. 200 subject data of 100 each diabetic and non-diabetic were evaluated. Complete procedure was kept very simple and free from the involvement of any iridologist. From the normalized iris, the region of interest was cropped. All 63 features were extracted using statistical, texture analysis, and two-dimensional discrete wavelet transformation. A comparison of accuracies of six different classifiers has been presented. The result shows 89.66% accuracy by the random forest classifier.
Abstract: Application of five implementations of three data mining classification techniques was experimented for extracting important insights from tourism data. The aim was to find out the best performing algorithm among the compared ones for tourism knowledge discovery. Knowledge discovery process from data was used as a process model. 10-fold cross validation method is used for testing purpose. Various data preprocessing activities were performed to get the final dataset for model building. Classification models of the selected algorithms were built with different scenarios on the preprocessed dataset. The outperformed algorithm tourism dataset was Random Forest (76%) before applying information gain based attribute selection and J48 (C4.5) (75%) after selection of top relevant attributes to the class (target) attribute. In terms of time for model building, attribute selection improves the efficiency of all algorithms. Artificial Neural Network (multilayer perceptron) showed the highest improvement (90%). The rules extracted from the decision tree model are presented, which showed intricate, non-trivial knowledge/insight that would otherwise not be discovered by simple statistical analysis with mediocre accuracy of the machine using classification algorithms.
Abstract: Developing complete mechanistic models for polymerization reactors is not easy, because complex reactions occur simultaneously; there is a large number of kinetic parameters involved and sometimes the chemical and physical phenomena for mixtures involving polymers are poorly understood. To overcome these difficulties, empirical models based on sampled data can be used instead, namely regression methods typical of machine learning field. They have the ability to learn the trends of a process without any knowledge about its particular physical and chemical laws. Therefore, they are useful for modeling complex processes, such as the free radical polymerization of methyl methacrylate achieved in a batch bulk process. The goal is to generate accurate predictions of monomer conversion, numerical average molecular weight and gravimetrical average molecular weight. This process is associated with non-linear gel and glass effects. For this purpose, an adaptive sampling technique is presented, which can select more samples around the regions where the values have a higher variation. Several machine learning methods are used for the modeling and their performance is compared: support vector machines, k-nearest neighbor, k-nearest neighbor and random forest, as well as an original algorithm, large margin nearest neighbor regression. The suggested method provides very good results compared to the other well-known regression algorithms.
Abstract: Accurate software reliability prediction not only enables developers to improve the quality of software but also provides useful information to help them for planning valuable resources. This paper examines the performance of three well-known data mining techniques (CART, TreeNet and Random Forest) for predicting software reliability. We evaluate and compare the performance of proposed models with Cascade Correlation Neural Network (CCNN) using sixteen empirical databases from the Data and Analysis Center for Software. The goal of our study is to help project managers to concentrate their testing efforts to minimize the software failures in order to improve the reliability of the software systems. Two performance measures, Normalized Root Mean Squared Error (NRMSE) and Mean Absolute Errors (MAE), illustrate that CART model is accurate than the models predicted using Random Forest, TreeNet and CCNN in all datasets used in our study. Finally, we conclude that such methods can help in reliability prediction using real-life failure datasets.
Abstract: Human skin detection recognized as the primary step in most of the applications such as face detection, illicit image filtering, hand recognition and video surveillance. The performance of any skin detection applications greatly relies on the two components: feature extraction and classification method. Skin color is the most vital information used for skin detection purpose. However, color feature alone sometimes could not handle images with having same color distribution with skin color. A color feature of pixel-based does not eliminate the skin-like color due to the intensity of skin and skin-like color fall under the same distribution. Hence, the statistical color analysis will be exploited such mean and standard deviation as an additional feature to increase the reliability of skin detector. In this paper, we studied the effectiveness of statistical color feature for human skin detection. Furthermore, the paper analyzed the integrated color and texture using eight classifiers with three color spaces of RGB, YCbCr, and HSV. The experimental results show that the integrating statistical feature using Random Forest classifier achieved a significant performance with an F1-score 0.969.
Abstract: As smartphones are equipped with various sensors,
there have been many studies focused on using these sensors to create
valuable applications. Human activity recognition is one such
application motivated by various welfare applications, such as the
support for the elderly, measurement of calorie consumption, lifestyle
and exercise patterns analyses, and so on. One of the challenges one
faces when using smartphone sensors for activity recognition is that
the number of sensors should be minimized to save battery power. In
this paper, we show that a fairly accurate classifier can be built that
can distinguish ten different activities by using only a single sensor
data, i.e., the smartphone accelerometer data. The approach that we
adopt to deal with this twelve-class problem uses various methods.
The features used for classifying these activities include not only the
magnitude of acceleration vector at each time point, but also the
maximum, the minimum, and the standard deviation of vector
magnitude within a time window. The experiments compared the
performance of four kinds of basic multi-class classifiers and the
performance of four kinds of ensemble learning methods based on
three kinds of basic multi-class classifiers. The results show that
while the method with the highest accuracy is ECOC based on
Random forest.
Abstract: Pulmonary Function Tests are important non-invasive
diagnostic tests to assess respiratory impairments and provides
quantifiable measures of lung function. Spirometry is the most
frequently used measure of lung function and plays an essential role
in the diagnosis and management of pulmonary diseases. However,
the test requires considerable patient effort and cooperation,
markedly related to the age of patients resulting in incomplete data
sets. This paper presents, a nonlinear model built using Multivariate
adaptive regression splines and Random forest regression model to
predict the missing spirometric features. Random forest based feature
selection is used to enhance both the generalization capability and the
model interpretability. In the present study, flow-volume data are
recorded for N= 198 subjects. The ranked order of feature importance
index calculated by the random forests model shows that the
spirometric features FVC, FEF25, PEF, FEF25-75, FEF50 and the
demographic parameter height are the important descriptors. A
comparison of performance assessment of both models prove that, the
prediction ability of MARS with the `top two ranked features namely
the FVC and FEF25 is higher, yielding a model fit of R2= 0.96 and
R2= 0.99 for normal and abnormal subjects. The Root Mean Square
Error analysis of the RF model and the MARS model also shows that
the latter is capable of predicting the missing values of FEV1 with a
notably lower error value of 0.0191 (normal subjects) and 0.0106
(abnormal subjects) with the aforementioned input features. It is
concluded that combining feature selection with a prediction model
provides a minimum subset of predominant features to train the
model, as well as yielding better prediction performance. This
analysis can assist clinicians with a intelligence support system in the
medical diagnosis and improvement of clinical care.
Abstract: Precipitation forecast is important in avoid incident of natural disaster which can cause loss in involved area. This review paper involves three techniques from artificial intelligence namely logistic regression, decisions tree, and random forest which used in making precipitation forecast. These combination techniques through VAR model in finding advantages and strength for every technique in forecast process. Data contains variables from rain domain. Adaptation of artificial intelligence techniques involved on rain domain enables the process to be easier and systematic for precipitation forecast.