Predictive Machine Research Articles

Analysing Health Insurance Customer Dataset to Determine Cross-Selling Potential

Health insurance cross-selling refers to the practice of offering additional or complementary insurance products to existing policyholders. Insurance providers leverage cross-selling, offering customers additional policies like dental or life insurance when they already have a basic health insurance plan. This study is conducted to focus on the application of machine learning techniques to predict health insurance cross-selling opportunities among South African customers. The aim of this study is to develop a cross-selling predictive machine learning model that can assist health insurance companies to identify potential customers for cross-selling probabilities. To achieve this goal, a quantitative research methodology is adopted, focusing on extracting a comprehensive dataset of health insurance consumer information and employing various machine learning algorithms using the Python programming language, including Random Forest, K-Nearest Neighbours, XGBoost classifier, and Logistic Regression algorithms to build the cross-selling predictive machine learning model. The experimental results obtained demonstrate the accuracy scores of four different machine learning algorithms trained using 1,000,000 customer dataset with 17 features, logistic regression is considered as the top-performing model. It achieved an accuracy score of 0.83 and an F1 score of 0.91. The analysis indicates that customers aged 30-60, with prior insurance, and longer service history are more likely to buy additional health insurance products. The findings of this research can help health insurers boost revenue by improving their customer targeting and retention strategies.

Optimized deep extreme learning machine for traffic prediction and autonomous vehicle lane change decision-making

Deep learning occupies a prominent position in the field of machine learning, where researchers have introduced autoencoder into extreme learning machine (ELM) to establish deep extreme learning machine (DELM), with the aim of enabling autonomous feature learning from data. This approach has found widespread applications in various domains, including finance and electrochemistry. However, the network accuracy of DELM is sensitive to the initial weighting and bias configurations, which can be negatively impacted by random initialization. To mitigate the adverse effects of such random initialization on network performance, this study employs two swarm intelligence optimization algorithms, particle swarm optimization (PSO) and harris hawks optimization (HHO), to optimize the input weights and biases of DELM. In addition, we apply the variational mode decomposition (VMD) technique to decompose the original data, enabling a deeper exploration of latent information within the data. Subsequently, two integrated models are constructed: VMD-HHO-DELM and VMD-PSO-DELM. We apply these models to address road traffic condition prediction and autonomous lane change decision-making problems, with a specific focus on the transportation domain. Through extensive experimentation on two real-world datasets, we observe that deep learning models based on swarm intelligence optimization algorithms exhibit excellent performance across various scenarios. In the context of road traffic condition prediction, our experiments reveal superior results under different time intervals and traffic volumes compared to traditional models, all while maintaining higher operational efficiency. For the autonomous lane change decision experiment, we establish a driver dissatisfaction model that aids in formulating more reasonable lane change decisions, thus reducing the frequency of abrupt lane changes.

Individualized Prediction of Outcomes of Hematopoietic Cell Transplantation for Sickle Cell Disease: A Machine Learning Approach

Complications of Sickle cell disease (SCD) are ameliorated by modifying therapies but, hematopoietic cell transplantation (HCT) remains the only therapeutic option with curative intent. The therapeutic dilemma faced by families regarding HCT is partly related to the uncertainty regarding outcomes and adverse events. Computational machine-learning (ML) methods add to population-level outcomes by helping determine generalizable predictive patterns and quantifying uncertainty in those estimates. The incorporation of uncertainty as well as predictions allows clinicians to assess the confidence level of the predictions, avoid over-reliance on potentially incorrect results, allows for risk management and mitigation by considering the potential range of outcomes. Thus, such models may permit better decision-making by considering different scenarios and their probabilities, and increase transparency and accountability, by providing insight into how the model arrived at its predictions. Our objective was to develop a personalized, predictive machine learning model derived from center for international bone marrow transplant research (CIBMTR)datasets which meets acceptable AUC, calibration slope and discrimination. Methods We applied ML methods to clinical parameters and both categorical and time-to-event outcomes in a de-identified CIBMTR dataset of patients undergoing HCT. A supervised random forest model was created with baseline covariates as independent variables. Model selection was performed by both the clinician and data scientists to create a model using prognostically relevant variables. Since the number and percentage of negative outcomes in HCT for SCD is smaller than the positive outcomes, the model is imbalanced and biased towards predicting positive outcomes. To counter the imbalance, we constructed a training dataset taking each outcome variable of interest, and included randomly sampled positive outcomes, typically 1.5-3 times the total instances of the variable of interest. We ran the test dataset, and used a random forest on 20 such trials. To account for the effect of the undersampling, we propose a positive threshold δ, and assigned a final prediction of a negative outcome if the average sum for an element is greater than delta. We performed a 2 times Repeated 10 Fold Cross Validation, to demonstrate our model's versatility and response to unknown data. The accuracy score may be misleading as it may result from the model being able to correctly predict the numerical majority of the positive outcomes, and does not indicate the ability to detect negative outcomes. We therefore estimated balanced accuracy which is the arithmetic mean of sensitivity and specificity. Thus, a higher balanced accuracy results only when both the negative and positive outcomes are predicted with a high accuracy. We also measured Area under the Receiver Operator Characteristic Curve (ROC AUC), a measure of the ability of a binary classifier to distinguish between classes. We define model confidence as the average probability across 20 trials, as well as our confidence in the model. We describe predictive probability percentage as : model confidence*our confidence in the model*100. Results We examined de-identified records of 1641 patients who underwent HCT and were reported to the CIBMTR. Patients were followed for a median of 47.8 months (0.3-312.9) Patient characteristics included 73.4% patient's age at HCT <18 years, Karnofsky-Lansky (KPS) score ≥ 90 in 74.7%, overall survival 91.2%, event-free survival 75.5%, graft failure (GF) 17.9%, AGVHD 18.3%, and CGVHD 22.3%. Predictive model performance is described in Table 1. Predictive variables that made a significant contribution, and predicted outcomes in three hypothetical scenarios are described in Figure 1. Overall, the predictive model provided acceptable AUC, Balanced accuracy, positive predictive value and sensitivity. Conclusions We report the development, testing, and validation of an ML model for individualized prediction of outcomes of HCT for SCD. The model provides acceptable AUC, accuracy, balanced accuracy, positive predictive value and sensitivity. This predictive model has the potential to aid clinicians in making shared decisions with their patients regarding HCT for SCD.

Discovery and Development of CD70 As a Cellular Therapy Target in High Risk Multiple Myeloma

Background: Even in the BCMA CAR-T era, multiple myeloma (MM) patients with high-risk genotypes have the poorest outcomes. We thus leveraged MMRF CoMMpass data to identify potential novel surface immunotherapeutic targets in high-risk MM. As described below, we focused on CD70, an immunotherapeutic target known to be upregulated on many hematologic and solid tumors but minimally expressed on normal tissue. While CD70 was previously explored in MM (McEarchern et al, Clin Cancer Res (2008)), it was not further investigated due to heterogeneous expression in newly diagnosed samples. Here we revisit this target to suggest cellular therapies against CD70 could be highly beneficial in high-risk MM patients, currently the greatest unmet need in the field. Methods: Primary patient samples were obtained under an IRB-approved protocol. CAR-T lentiviral transduction was performed on paired CD4/CD8 human T cells from at least 3 healthy donors. Effector function was assessed by cytotoxicity against MM cell lines and CAR-T profiling for memory and exhaustion markers. In vivo efficiency was assessed using bioluminescence. ATAC-seq peaks were called using Promo and HOMER (data from Jin et al. Blood (2018)). Machine learning was performed using XGBoost approach applied to CoMMpass RNA-seq data with 80% of samples for training and 20% as test set. Results: Analyzing CoMMpass mRNA data (release IA19) we first evaluated surface protein encoding-genes upregulated in tumors at relapse compared to diagnosis. Among >100 significantly upregulated genes, we focused on CD70 given known promise as an immunotherapy target. Further analysis across genomic subtypes revealed a specific, marked upregulation of CD70 in t(4;14) and gain 1q ( Fig. 1), as well as R-ISS 3 at diagnosis. Flow cytometry profiling of 8 cell line models, including t(4;14) models KHM11, KMS26, and LP1, revealed consistent CD70 protein expression with antigen density similar to BCMA. Profiling of relapsed primary MM samples confirmed that 19 of 28 samples had CD138+ plasma cells positive for CD70 expression. Of 5 patients in whom both BCMA and CD70 were measured, antigen densities were also similar. To target CD70, we then designed 10 unique anti-CD70 CAR-T constructs. 9 of these were scFv-based binders derived from therapeutic antibody fragments currently in preclinical or clinical development. By a structure-guided design based on the CD70:CD27 co-crystal structure (PDB: 7KX0), we also derived a unique truncated fragment of CD27 as an alternative CD70 binding element. All CD70 CARs were efficacious in in vitro cytotoxicity assays targeting MM cell lines AMO1, MM1.S, LP1, and RPMI8226. We then selected the top 3 performing scFv based-CARs as well as our natural ligand (CD27) CAR-T for in vivo studies. All 4 CD70 CAR-Ts prolonged survival of MM.1S-bearing mice compared to empty CAR ( n=5 mice/arm, p<0.005 by log-rank), similar to BCMA CAR-T controls. Remarkably, the CD27-based CAR showed superior expansion (50-85x, p< 0.0001 by ANOVA) and persistence compared to scFv-based ( Fig. 2). These results underscore that targeting CD70 using a natural ligand-based CAR-T represents a promising preclinical candidate for treatment of relapsed and high risk MM. We further investigated why CD70 may be upregulated in high risk MM. Analysis of ChIP-seq in ENCODE and primary MM tumor ATAC-seq was used to nominate a set of transcription factors that bind the CD70 locus. Extending from CoMMpass data, we built a predictive machine learning model for CD70 expression as a function of transcription factor gene expression, nominating key transcription factors TFAP2A, LEF1, MYB, and PAX5 as CD70 regulators; validation is now underway. Additionally, we showed that knockout of NSD2 in t(4;14) models KMS34 and KMS11 led to significantly decreased CD70 surface expression, implicating this central epigenetic driver in CD70 overexpression in this high-risk MM subtype. Finally, we also demonstrated that treatment with the hypomethylating agent azacytidine can drive increased CD70 expression, similar to that shown in AML models. Conclusion: Taken together, targeting CD70 appears to be a highly promising cellular therapeutic strategy for MM, with particularly high utility in high-risk patients after BCMA CAR-T relapse. We are continuing preclinical and future clinical development of this CAR-T strategy, using our natural ligand CD27 design with extended in vivo persistence.

The predictive machine learning model of a hydrated inverse vulcanized copolymer for effective mercury sequestration from wastewater

Inverse vulcanized polysulfides (IVP) are promising sulfur-enriched copolymers with unconventional properties irresistible for diverse applications like Hg2+ remediation. Nevertheless, due to their inherent hydrophobic nature, these copolymers still offer low Hg2+ uptake capacity. Herein, we reported the synthesis of IVP by reacting molten sulfur with 4-vinyl benzyl chloride, followed by their functionalization using N-methyl D-glucamine (NMDG) to increase the hydration of the developed IVP. The chemical composition and structure of the functionalized IVP were proposed based on FTIR and XPS analysis. The functionalized IVP demonstrated a high mercury adsorption capacity of 608 mg/g (compared to <26 mg/g for common IVP) because of rich sulfur and hydrophilic regions. NMDG functionalized IVP removed 100 % Hg2+ from a low feed concentration (10–50 mg/l). A predictive machine learning model was also developed to predict the amount of mercury removed (%) using GPR, ANN, Decision Tree, and SVM algorithms. Hyperparameter and loss function optimization was also carried out to reduce the prediction error. The optimized GPR algorithm demonstrated high R2 (0.99 (training) and 0.98 (unseen)) and low RMSE (2.74 (training) and 2.53 (unseen)) values indicating its goodness in predicting the amount of mercury removed. The produced functionalized IVP can be regenerated and reused with constant Hg2+ uptake capacity. Sulfur is the waste of the petrochemical industry and is abundantly available, making the functionalized IVP a sustainable and cheap adsorbent that can be produced for high-volume Hg2+ remediation. Environmental implicationThis research effectively addresses the removal of the global top-priority neurotoxic pollutant mercury, which is toxic even at low concentrations. We attempted to remove the Hg2+ utilizing an inexpensive adsorbent developed by NMDG functionalized copolymer of molten sulfur and VBC. A predictive machine learning model was also formulated to predict the amount of mercury removal from wastewater with only a 0.05 % error which shows the goodness of the developed model. This work is critical in utilizing this low-cost adsorbent and demonstrates its potential for large-scale industrial application.

Improving serious illness conversations in oncology: A machine learning approach that integrates natural language processing for mortality prediction.

590 Background: ML-based mortality prediction tools in oncology can optimize clinical decisions and prompt end-of-life care discussions. Patients with advanced cancer who have engaged in Goals of Care (GoC) conversations report improved quality of life and better care alignment. However, oncologists often have overly optimistic prognoses and miss timely GoC discussions. Clinical notes are a valuable source of information, but processing and extracting data from them is time-consuming and labor-intensive. To address this issue, we have developed a machine learning application that ingests clinical notes and structured data from electronic health records (EHRs) to generate a 180-day mortality risk, prompting oncologists for GoC conversations. Methods: A predictive machine learning model was developed using data from cancer patients aged 21 and above, diagnosed between January 2016 and December 2021. Data was collected from various sources, including cancer and death registry and the EHR. By analyzing structured and unstructured data from ambulatory progress notes, a clinical profile was created for each patient. The model utilized Spark-NLP for preprocessing, applying word2vec embedding and pre-trained NER models to extract information on diseases, symptoms, procedures, treatments, and medications. Feature engineering techniques were used to select the best NLP features, combined with structured data. The model was trained using 894 patients, employing Random Forest Classifier with 10-fold cross-validation, and tested on a separate set of 43,274 patients. Performance evaluation included ROC AUC, PR AUC, and F1 Score metrics. Results: After the fine tuning, the best model showed an AUC-ROC of 0.88 on the train set and 0.75 on the test set. At a threshold of 0.44, the model achieved a balanced performance with a sensitivity of 0.70 and specificity of 0.71 on the testing set. Conclusions: Our team pioneered the development of an automated multi-modality pipeline that combines unstructured real-world data with structured data, allowing for training and testing of a fusion model. This automation opens doors for scaling and dissemination, to enhance mortality prediction. Future works will involve qualitative analysis of implementation and acceptance in clinical practice.

The Experience of Meaningful Coincidences Is Associated with Stronger Alpha Power Increases during an Eyes-closed Resting Condition: A Bayesian Replication Approach.

Recognizing and perceiving meaningful patterns in an ever-changing environment is fundamental to (human) beings. Apophenia, patternicity, and the propensity to perceive meaningful coincidences might result from the human brain working as a prediction machine that constantly matches sensory information to prior expectations. The propensity for Type I errors varies between people and, at its extreme, is associated with symptoms of schizophrenia. However, on a nonclinical level seeing meaning in randomness might be benevolent and was found to be associated with creativity and openness. However, hardly any neuroscientific investigation has examined EEG patterns of the propensity to experience meaningful coincidences in this manner. We hypothesized deviations in brain functions as one potential reason why some people experience more meaning in random arrangements than others. The gating by inhibition theory suggests that alpha power increases represent basic control mechanisms of sensory processes during varying task requirements. We found that people perceiving more meaningful coincidences had higher alpha power during an eyes-closed versus eyes-opened condition compared with people experiencing less meaningful coincidences. This indicates deviations in the sensory inhibition mechanism of the brain, which are critically relevant for higher cognitive functions. Applying Bayesian statistics, we replicated this finding in another independent sample.

S207 Predictive Machine Learning Models for Recurrent C. difficile Infection in Inflammatory Bowel Disease Patients Based on First-Episode CDI Treatment

S207 Predictive Machine Learning Models for Recurrent C. difficile Infection in Inflammatory Bowel Disease Patients Based on First-Episode CDI Treatment

Global mangrove soil organic carbon stocks dataset at 30 m resolution for the year 2020 based on spatiotemporal predictive machine learning.

This dataset presents global soil organic carbon stocks in mangrove forests at 30 m resolution, predicted for 2020. We used spatiotemporal ensemble machine learning to produce predictions of soil organic carbon content and bulk density (BD) to 1 m soil depth, which were then aggregated to calculate soil organic carbon stocks. This was done by using training data points of both SOC (%) and BD in mangroves from a global dataset and from recently published studies, and globally consistent predictive covariate layers. A total of 10,331 soil samples were validated to have SOC (%) measurements and were used for predictive soil mapping. We used time-series remote sensing data specific to time periods when the training data were sampled, as well as long-term (static) layers to train an ensemble of machine learning model. Ensemble models were used to improve performance, robustness and unbiasedness as opposed to just using one learner. In addition, we performed spatial cross-validation by using spatial blocking of training data points to assess model performance. We predicted SOC stocks for the 2020 time period and applied them to a 2020 mangrove extent map, presenting both mean predictions and prediction intervals to represent the uncertainty around our predictions. Predictions are available for download under CC-BY license from 10.5281/zenodo.7729491 and also as Cloud-Optimized GeoTIFFs (global mosaics).

Evaluation of Deep Learning Algorithms for KASE Index Prediction

Security indexes are the primary instruments that are used to analyze the circumstances of the financial market. A considerable proportion of a country’s total wealth is held in the form of investments in its stock market. If future stock market patterns can be accurately predicted, it may be possible for investors to optimize the returns they receive on their investments. Yet, as a result of nonlinearity and nonstationarity, it is difficult to accurately forecast financial data. The purpose of this investigation is to determine how well predictive machine learning models perform in relation to the stock market. This analysis makes use of daily close price data for the KASE index on the Kazakhstan Stock Exchange from March 2018 all the way through March 2023. In order to make accurate predictions, we use a total of four different machine learning algorithms. According to the findings, the deep learning model fares far better than the competition in terms of the accuracy of its predictions. In addition, the neural network approach and the random forest method are considered to be the best, while the support vector regression method is placed second. Criteria for evaluation include mean absolute error (also known as MAE), mean square error (also known as MSE), and root mean square error (RMSE).

Federated learning enabled graph convolutional autoencoder and factorization machine for potential friendship prediction in social networks

Friendships are the keystone of social networks. Predicting potential friendships of users in social networks has become a critical task in the real world. However, the computational models proposed by previous researchers do not effectively capture the behavior preferences of users, which limits the recommendation results. Moreover, the process of generating predicted values has not considered the cross-information among different features. Therefore, we design a potential friendship prediction model based on graph convolutional autoencoder and factorization machine (GCAFM). The GCAFM model uses a graph convolutional autoencoder to learn the hidden feature of users, which can make the similar users as close as possible in the embedding space. The hidden feature of the target users is then subjected to an element-to-element multiplication operation, and the product results are subsequently inputting into a factorization machine model that capable of feature crossover to get the predicted friendship. In addition, we use the federated learning framework to train the GCAFM model, which ensures the privacy of social network data. The experimental results on Douban and FilmTrust social network datasets show that the GCAFM model exceeds four comparison models under various kinds of evaluation metrics. The GCAFM model scored 0.856 and 0.850 under the AUC metric, 0.683 and 0.702 under the AUPR metric, 0.805 and 0.844 under the Precision metric on Douban and FilmTrust datasets.

Predictive machine learning models for optimization of direct solar steam generation

Direct solar steam generation (DSSG) has gained significant consideration in the recent decade because of its ability to generate freshwater, relying on renewable solar energy. Despite experimental data abundance, it is still difficult to optimize DSSG under certain conditions regarding fluid surface temperature changes (Ttop) and evaporation efficiency (η). This study investigates six predictive machine learning models, including multilayer perceptron (MLP), support vector regression (SVR), decision tree (DT), random forest (RF), adaptive boosting ensemble (ADA-BE), and combinations of them, to model Ttop and η in interfacial and volumetric DSSG systems. The models are trained on experimental data, and their performance is evaluated using various metrics. Based on the findings of the study, the DT (total R2 = 0. 9900) and DT-SVR combo (total R2 = 0.9829) are the best models to predict η in interfacial and volumetric systems, respectively. Results show that interfacial DT-MLP combo (total R2 = 0.9964) and volumetric DT-ADA-BE (total R2 = 0.9870) models predict Ttop more accurately. The study predicts that the ηmax of 85 ± 5 % and 90.91 ± 5 % will be obtained under one sun (1 kW/m2) using GNP-MWCNT with 0.015 weight percentage in volumetric and using Au-HT-wood with a thickness of 14.78 mm in interfacial approaches, respectively.

Framing Machine Learning Opportunities for Hypotension Prediction in Perioperative Care: A Socio-technical Perspective

Hypotension during perioperative care, if undetected or uncontrolled, can lead to serious clinical complications. Predictive machine learning models, based on routinely collected EHR data, offer potential for early warning of hypotension to enable proactive clinical intervention. However, while research has demonstrated the feasibility of such machine learning models, little effort is made to ground their formulation and development in socio-technical context of perioperative care work. To address this, we present a study of collaborative work practices of clinical teams during and after surgery with specific emphasis on the organisation of hypotension management. The findings highlight where predictive insights could be usefully deployed to reconfigure care and facilitate more proactive management of hypotension. We further explore how the socio-technical insights help define key parameters of machine learning prediction tasks to align with the demands of collaborative clinical practice. We discuss more general implications for the design of predictive machine learning in hospital care.

AI-Driven Decision Support for Early Detection of Cardiac Events: Unveiling Patterns and Predicting Myocardial Ischemia.

Cardiovascular diseases (CVDs) account for a significant portion of global mortality, emphasizing the need for effective strategies. This study focuses on myocardial infarction, pulmonary thromboembolism, and aortic stenosis, aiming to empower medical practitioners with tools for informed decision making and timely interventions. Drawing from data at Hospital Santa Maria, our approach combines exploratory data analysis (EDA) and predictive machine learning (ML) models, guided by the Cross-Industry Standard Process for Data Mining (CRISP-DM) methodology. EDA reveals intricate patterns and relationships specific to cardiovascular diseases. ML models achieve accuracies above 80%, providing a 13 min window to predict myocardial ischemia incidents and intervene proactively. This paper presents a Proof of Concept for real-time data and predictive capabilities in enhancing medical strategies.

Machine and Deep Learning Algorithms for COVID-19 Mortality Prediction Using Clinical and Radiomic Features

Aim: Machine learning (ML) and deep learning (DL) predictive models have been employed widely in clinical settings. Their potential support and aid to the clinician of providing an objective measure that can be shared among different centers enables the possibility of building more robust multicentric studies. This study aimed to propose a user-friendly and low-cost tool for COVID-19 mortality prediction using both an ML and a DL approach. Method: We enrolled 2348 patients from several hospitals in the Province of Reggio Emilia. Overall, 19 clinical features were provided by the Radiology Units of Azienda USL-IRCCS of Reggio Emilia, and 5892 radiomic features were extracted from each COVID-19 patient’s high-resolution computed tomography. We built and trained two classifiers to predict COVID-19 mortality: a machine learning algorithm, or support vector machine (SVM), and a deep learning model, or feedforward neural network (FNN). In order to evaluate the impact of the different feature sets on the final performance of the classifiers, we repeated the training session three times, first using only clinical features, then employing only radiomic features, and finally combining both information. Results: We obtained similar performances for both the machine learning and deep learning algorithms, with the best area under the receiver operating characteristic (ROC) curve, or AUC, obtained exploiting both clinical and radiomic information: 0.803 for the machine learning model and 0.864 for the deep learning model. Conclusions: Our work, performed on large and heterogeneous datasets (i.e., data from different CT scanners), confirms the results obtained in the recent literature. Such algorithms have the potential to be included in a clinical practice framework since they can not only be applied to COVID-19 mortality prediction but also to other classification problems such as diabetic prediction, asthma prediction, and cancer metastases prediction. Our study proves that the lesion’s inhomogeneity depicted by radiomic features combined with clinical information is relevant for COVID-19 mortality prediction.

Individual prediction tendencies do not generalize across modalities.

Predictive processing theories, which model the brain as a "prediction machine", explain a wide range of cognitive functions, including learning, perception and action. Furthermore, it is increasingly accepted that aberrant prediction tendencies play a crucial role in psychiatric disorders. Given this explanatory value for clinical psychiatry, prediction tendencies are often implicitly conceptualized as individual traits or as tendencies that generalize across situations. As this has not yet explicitly been shown, in the current study, we quantify to what extent the individual tendency to anticipate sensory features of high probability generalizes across modalities. Using magnetoencephalography (MEG), we recorded brain activity while participants were presented with a sequence of four different (either visual or auditory) stimuli, which changed according to predefined transitional probabilities of two entropy levels: ordered vs. random. Our results show that, on a group-level, under conditions of low entropy, stimulus features of high probability are preactivated in the auditory but not in the visual modality. Crucially, the magnitude of the individual tendency to predict sensory events seems not to correlate between the two modalities. Furthermore, reliability statistics indicate poor internal consistency, suggesting that the measures from the different modalities are unlikely to reflect a single, common cognitive process. In sum, our findings suggest that quantification and interpretation of individual prediction tendencies cannot be generalized across modalities.

Hybrid LSTM Model to Predict the Level of Air Pollution in Montenegro

Air pollution is a critical environmental concern that poses significant health risks and affects multiple aspects of human life. ML algorithms provide promising results for air pollution prediction. In the existing scientific literature, Long Short-Term Memory (LSTM) predictive models, as well as their combination with other statistical and machine learning approaches, have been utilized for air pollution prediction. However, these combined algorithms may not always provide suitable results due to the stochastic nature of the factors that influence air pollution, improper hyperparameter configurations, or inadequate datasets and data characterized by great variability and extreme dispersion. The focus of this paper is applying and comparing the performance of Support Vector Machine and hybrid LSTM regression models for air pollution prediction. To identify optimal hyperparameters for the LSTM model, a hybridization with the Genetic Algorithm is proposed. To mitigate the risk of overfitting, the bagging technique is employed on the best LSTM model. The proposed predicitive model aims to determine the Common Air Quality Index level for the next hour in Niksic, Montenegro. With the hybridization of the LSTM algorithm and by applying the bagging technique, our approach aims to significantly enhance the accuracy and reliability of hourly air pollution prediction. The major contribution of this paper is in the application of advanced machine learning analysis and the combination of the LSTM, Genetic Algorithm, and bagging techniques, which have not been previously employed in the analysis of air pollution in Montenegro. The proposed model will be made available to interested management structures, local governments, national entities, or other relevant institutions, empowering them to make effective pollution level predictions and take appropriate measures.

A Novel Predictive Machine Learning Model Integrating Cytokines in Cervical-Vaginal Mucus Increases the Prediction Rate for Preterm Birth.

Preterm birth (PB) is a leading cause of perinatal morbidity and mortality. PB prediction is performed by measuring cervical length, with a detection rate of around 70%. Although it is known that a cytokine-mediated inflammatory process is involved in the pathophysiology of PB, none screening method implemented in clinical practice includes cytokine levels as a predictor variable. Here, we quantified cytokines in cervical-vaginal mucus of pregnant women (18-23.6 weeks of gestation) with high or low risk for PB determined by cervical length, also collecting relevant obstetric information. IL-2, IL-6, IFN-γ, IL-4, and IL-10 were significantly higher in the high-risk group, while IL-1ra was lower. Two different models for PB prediction were created using the Random Forest machine-learning algorithm: a full model with 12 clinical variables and cytokine values and the adjusted model, including the most relevant variables-maternal age, IL-2, and cervical length- (detection rate 66 vs. 87%, false positive rate 12 vs. 3.33%, false negative rate 28 vs. 6.66%, and area under the curve 0.722 vs. 0.875, respectively). The adjusted model that incorporate cytokines showed a detection rate eight points higher than the gold standard calculator, which may allow us to identify the risk PB risk more accurately and implement strategies for preventive interventions.

Forecasting next-hour electricity demand in small-scale territories: Evidence from Jordan

In exceptional times of wars, natural crises (e.g., snow storms), or hosting massive events (e.g., international sports events), prior knowledge of hour-by-hour electricity demand might become critical for the concerned areas. Anticipating the next-hour demand within a bounded location might help cope with challenging situations. In this paper, we are concerned with the problem of electricity demand forecasting for the next hour within a small area in the context of the country of Jordan relying exclusively on historical electricity consumption. Existing electricity demand forecasting models predict electricity demand for the next day for the whole country of Jordan or for large cities using numerous indicators such as weather conditions and season-related features, together with historical electricity consumption. Our proposed solution consists of preprocessing the hourly-electricity-consumption data to rearrange it into time series of 25-hour length. Then, we applied supervised machine learning algorithms to build predictive models for electricity demand. Our methods proved effective to a certain degree in forecasting electricity demand using time series of historical electricity consumptions only. We employ several predictive machine learning models: neural network, decision tree, random forest, AdaBoost, support vector regression, nearest neighbors, linear regression, bayesian ridge, partial least squares, gradient boosting, and principal component regression. We found that the most critical indicators in our predictive models are the electricity consumption of the two hours preceding the target hour and the electricity consumption the day before in the hour (the same as the target hour) and the next hour. Our predictive models can efficiently forecast electricity demand in a small rural area in Jordan with an accuracy of nearly 97% at best and 92% at least. Our predictive models are constructed using historical data that includes the electricity consumption of a small Jordanian district for the years 2019–2021.

Advancing the comprehension of consumer price index and influencing factors: insight into the mechanism based on prediction machine learning models

The consumer price index (CPI) is an important indicator to measuring inflation or deflation. Its changes are closely related to residents' lives, and also affect the direction of national macroeconomic policy formulation. In recent years, the combination of economics with mathematical models such as machine learning and macro-statistics has gradually become a hot topic. In this paper, the impact of different types of CPI on China's overall CPI was discussed. Machine learning prediction and correlation analysis of various types of influencing factors and CPI. The machine learning model of the regression decision tree process predicted CPI more accurately. Spearman correlation analysis showed that CPI was mainly positively related to goods and services, living, medical care, food, tobacco and alcohol, clothing and so on, while CPI was mainly negatively related to the transport and communication of residents.