Multivariate Time Series Datasets Research Articles

Multivariate data binning and examples generation to build a Diabetic Retinopathy classifier based on temporal clinical and analytical risk factors

In this paper, we explore the possibility of exploiting retrospective clinical data from Electronic Health Records (EHR) for classification tasks in chronic patients. The different intervals, short length and high class imbalance make it unfeasible to use traditional time series techniques. The first contribution of the paper is a preprocessing method to construct a multivariate time series dataset using EHR data, which infers missing data and regularizes the data frequency. The second contribution addresses class imbalance by using domain knowledge and existing short EHR series. We synthetically extrapolate patients’ data by using similar long time series and a fuzzy-based approach. The paper addresses the problem of detection of Diabetic Retinopathy (DR). Expert domain knowledge from ophthalmologists has been used in the proposed techniques to guide the processing of time series. The novelty in that case study consists in not using eye-fundus image analysis. Instead, the proposed methods are based solely on EHR data. Several multivariate multiclass time series classifiers are used to detect the four levels of DR severity from the pre-processed data sequences. Experiments prove the quality of the sequence preprocessing techniques proposed for EHR data. Results indicate that the TapNet classifier is the best one for DR grading. Despite being tested for DR detection, the proposed data preparation methods are applicable to other diseases with similar characteristics.

Forecasting Pitch Response of Floating Offshore Wind Turbines with a Deep Learning Model

The design and optimization of floating offshore wind turbines (FOWTs) pose significant challenges, stemming from the complex interplay among aerodynamics, hydrodynamics, structural dynamics, and control systems. In this context, this study introduces an innovative method for forecasting the dynamic behavior of FOWTs under various conditions by merging Convolutional Neural Network (CNN) with a Gated Recurrent Unit (GRU) network. This model outperforms traditional numerical models by delivering precise and efficient predictions of dynamic FOWT responses. It adeptly handles computational complexities and reduces processing duration, while maintaining flexibility and effectively managing nonlinear dynamics. The model’s prowess is showcased through an analysis of a spar-type FOWT in a multivariate parallel time series dataset using the CNN–GRU structure. The outcomes are notably promising, underscoring the model’s proficiency in accurately forecasting the performance of FOWTs.

Improved Ozone Level Detection through Feature Selection with Modified Whale Optimization Algorithm

This study presents a new approach for ozone level detection through feature selection by the modified Whale Optimization Algorithm (mWOA). This study aims to enhance the accuracy and efficiency of ozone level prediction models by selecting the most informative features from the dataset. As air quality deterioration poses significant risks to both human health and ecological equilibrium, pinpointing relevant features becomes essential for boosting prediction accuracy. The scope of the research includes comparing the performance of mWOA with the original WOA in two feature selection techniques: filter-based and wrapper-based. The experiments run proposed approaches on a multivariate time-series dataset with 20 repetitions. The evaluation criteria include processing time, number of features selected, and classification accuracy obtained by the kNN classifier. The statistical results demonstrate the effectiveness of the proposed mWOA approach, outperforming WOA due to the modified control parameter that enables a more precise exploration of the search area. The findings of this study reveal the improved performance of mWOA in selecting informative features, resulting in better prediction on average: 93.75% for filter-based and 94.49% for wrapper-based. In conclusion, the wrapper-based feature selection using the mWOA approach proves to be a valuable asset in enhancing the accuracy and efficiency of ozone level detection models. In the future, the proposed technique can be used for more applications in environmental science and engineering research.

Latent Diffusion Transformer for Probabilistic Time Series Forecasting

The probability prediction of multivariate time series is a notoriously challenging but practical task. This research proposes to condense high-dimensional multivariate time series forecasting into a problem of latent space time series generation, to improve the expressiveness of each timestamp and make forecasting more manageable. To solve the problem that the existing work is hard to extend to high-dimensional multivariate time series, we present a latent multivariate time series diffusion framework called Latent Diffusion Transformer (LDT), which consists of a symmetric statistics-aware autoencoder and a diffusion-based conditional generator, to implement this idea. Through careful design, the time series autoencoder can compress multivariate timestamp patterns into a concise latent representation by considering dynamic statistics. Then, the diffusion-based conditional generator is able to efficiently generate realistic multivariate timestamp values on a continuous latent space under a novel self-conditioning guidance which is modeled in a non-autoregressive way. Extensive experiments demonstrate that our model achieves state-of-the-art performance on many popular high-dimensional multivariate time series datasets.

Fully-Connected Spatial-Temporal Graph for Multivariate Time-Series Data

Multivariate Time-Series (MTS) data is crucial in various application fields. With its sequential and multi-source (multiple sensors) properties, MTS data inherently exhibits Spatial-Temporal (ST) dependencies, involving temporal correlations between timestamps and spatial correlations between sensors in each timestamp. To effectively leverage this information, Graph Neural Network-based methods (GNNs) have been widely adopted. However, existing approaches separately capture spatial dependency and temporal dependency and fail to capture the correlations between Different sEnsors at Different Timestamps (DEDT). Overlooking such correlations hinders the comprehensive modelling of ST dependencies within MTS data, thus restricting existing GNNs from learning effective representations. To address this limitation, we propose a novel method called Fully-Connected Spatial-Temporal Graph Neural Network (FC-STGNN), including two key components namely FC graph construction and FC graph convolution. For graph construction, we design a decay graph to connect sensors across all timestamps based on their temporal distances, enabling us to fully model the ST dependencies by considering the correlations between DEDT. Further, we devise FC graph convolution with a moving-pooling GNN layer to effectively capture the ST dependencies for learning effective representations. Extensive experiments show the effectiveness of FC-STGNN on multiple MTS datasets compared to SOTA methods. The code is available at https://github.com/Frank-Wang-oss/FCSTGNN.

Parallel multivariate deep learning models for time-series prediction: A comparative analysis in Asian stock markets

<span lang="EN-US">This study investigates deep learning models for financial data prediction and examines whether the architecture of a deep learning model and time-series data properties affect prediction accuracy. Comparing the performance of convolutional neural network (CNN), long short-term memory (LSTM), Stacked-LSTM, CNN-LSTM, and convolutional LSTM (ConvLSTM) when used as a prediction approach to a collection of financial time-series data is the main methodology of this study. In this instance, only those deep learning architectures that can predict multivariate time-series data sets in parallel are considered. This research uses the daily movements of 4 (four) Asian stock market indices from 1 January 2020 to 31 December 2020. Using data from the early phase of the spread of the Covid-19 pandemic that has created worldwide economic turmoil is intended to validate the performance of the analyzed deep learning models. Experiment results and analytical findings indicate that there is no superior deep learning model that consistently makes the most accurate predictions for all states' financial data. In addition, a single deep learning model tends to provide more accurate predictions for more stable time-series data, but the hybrid model is preferred for more chaotic time-series data.</span>

Universal representation learning for multivariate time series using the instance-level and cluster-level supervised contrastive learning

AbstractThe multivariate time series classification (MTSC) task aims to predict a class label for a given time series. Recently, modern deep learning-based approaches have achieved promising performance over traditional methods for MTSC tasks. The success of these approaches relies on access to the massive amount of labeled data (i.e., annotating or assigning tags to each sample that shows its corresponding category). However, obtaining a massive amount of labeled data is usually very time-consuming and expensive in many real-world applications such as medicine, because it requires domain experts’ knowledge to annotate data. Insufficient labeled data prevents these models from learning discriminative features, resulting in poor margins that reduce generalization performance. To address this challenge, we propose a novel approach: supervised contrastive learning for time series classification (SupCon-TSC). This approach improves the classification performance by learning the discriminative low-dimensional representations of multivariate time series, and its end-to-end structure allows for interpretable outcomes. It is based on supervised contrastive (SupCon) loss to learn the inherent structure of multivariate time series. First, two separate augmentation families, including strong and weak augmentation methods, are utilized to generate augmented data for the source and target networks, respectively. Second, we propose the instance-level, and cluster-level SupCon learning approaches to capture contextual information to learn the discriminative and universal representation for multivariate time series datasets. In the instance-level SupCon learning approach, for each given anchor instance that comes from the source network, the low-variance output encodings from the target network are sampled as positive and negative instances based on their labels. However, the cluster-level approach is performed between each instance and cluster centers among batches, as opposed to the instance-level approach. The cluster-level SupCon loss attempts to maximize the similarities between each instance and cluster centers among batches. We tested this novel approach on two small cardiopulmonary exercise testing (CPET) datasets and the real-world UEA Multivariate time series archive. The results of the SupCon-TSC model on CPET datasets indicate its capability to learn more discriminative features than existing approaches in situations where the size of the dataset is small. Moreover, the results on the UEA archive show that training a classifier on top of the universal representation features learned by our proposed method outperforms the state-of-the-art approaches.

Distributional Drift Adaptation With Temporal Conditional Variational Autoencoder for Multivariate Time Series Forecasting.

Due to the nonstationary nature, the distribution of real-world multivariate time series (MTS) changes over time, which is known as distribution drift. Most existing MTS forecasting models greatly suffer from distribution drift and degrade the forecasting performance over time. Existing methods address distribution drift via adapting to the latest arrived data or self-correcting per the meta knowledge derived from future data. Despite their great success in MTS forecasting, these methods hardly capture the intrinsic distribution changes, especially from a distributional perspective. Accordingly, we propose a novel framework temporal conditional variational autoencoder (TCVAE) to model the dynamic distributional dependencies over time between historical observations and future data in MTSs and infer the dependencies as a temporal conditional distribution to leverage latent variables. Specifically, a novel temporal Hawkes attention (THA) mechanism represents temporal factors that subsequently fed into feedforward networks to estimate the prior Gaussian distribution of latent variables. The representation of temporal factors further dynamically adjusts the structures of Transformer-based encoder and decoder to distribution changes by leveraging a gated attention mechanism (GAM). Moreover, we introduce conditional continuous normalization flow (CCNF) to transform the prior Gaussian to a complex and form-free distribution to facilitate flexible inference of the temporal conditional distribution. Extensive experiments conducted on six real-world MTS datasets demonstrate the TCVAE's superior robustness and effectiveness over the state-of-the-art MTS forecasting baselines. We further illustrate the TCVAE applicability through multifaceted case studies and visualization in real-world scenarios.

Correlation Recurrent Units: A Novel Neural Architecture for Improving the Predictive Performance of Time-Series Data.

Time-series forecasting (TSF) is a traditional problem in the field of artificial intelligence, and models such as recurrent neural network, long short-term memory, and gate recurrent units have contributed to improving its predictive accuracy. Furthermore, model structures have been proposed to combine time-series decomposition methods such as seasonal-trend decomposition using LOESS. However, this approach is learned in an independent model for each component, and therefore, it cannot learn the relationships between the time-series components. In this study, we propose a new neural architecture called a correlation recurrent unit (CRU) that can perform time-series decomposition within a neural cell and learn correlations (autocorrelation and correlation) between each decomposition component. The proposed neural architecture was evaluated through comparative experiments with previous studies using four univariate and four multivariate time-series datasets. The results showed that long- and short-term predictive performance was improved by more than 10%. The experimental results indicate that the proposed CRU is an excellent method for TSF problems compared to other neural architectures.

Adaptive Memory Networks With Self-Supervised Learning for Unsupervised Anomaly Detection

Unsupervised anomaly detection aims to build models to effectively detect unseen anomalies by only training on the normal data. Although previous reconstruction-based methods have made fruitful progress, their generalization ability is limited due to two critical challenges. First, the training dataset only contains normal patterns, which limits the model generalization ability. Second, the feature representations learned by existing models often lack representativeness which hampers the ability to preserve the diversity of normal patterns. In this paper, we propose a novel approach called Adaptive Memory Network with Self-supervised Learning (AMSL) to address these challenges and enhance the generalization ability in unsupervised anomaly detection. Based on the convolutional autoencoder structure, AMSL incorporates a self-supervised learning module to learn general normal patterns and an adaptive memory fusion module to learn rich feature representations. Experiments on four public multivariate time series datasets demonstrate that AMSL significantly improves the performance compared to other state-of-the-art methods. Specifically, on the largest CAP sleep stage detection dataset with 900 million samples, AMSL outperforms the second-best baseline by 4%+ in both accuracy and F1 score. Apart from the enhanced generalization ability, AMSL is also more robust against input noise.

A graph embedding based fault detection framework for process systems with multi-variate time-series datasets

Due to the enormous potential of modelling, graph-based approaches have been used for various applications in the process industries. In this study, we propose a fault detection framework through graphs by utilising its attributes in the form of node embeddings. Shallow embedding methods are deployed to generate node embedding vectors. Shallow embedding methods are broadly classified into matrix factorisation and skip-gram-based methods. Node2vec and Deepwalk fall under skip-gram models, while GraphRep and HOPE constitute the Matrix factorisation methods. Node embedding values generated from these methods are then fed to the variational auto-encoder, which ranks the nodes in reconstruction loss value. The node embedding reconstruction loss values exceeding a particular threshold are considered outliers. The proposed work has been validated on NPCIL power-flux data and the benchmark Tennessee Eastman data. The results indicate that skip-gram models, especially Node2vec-VAE, outperformed the matrix factorisation methods for both the above-mentioned datasets.

MultiBEATS: Blocks of eigenvalues algorithm for multivariate time series dimensionality reduction

Multivariate Time Series are sequences of observations taken from multiple sources. The proliferation of environments in which data is collected by means of sensors and the adoption of data-based services reveals the importance of their efficient analysis. In smart environments, the analysis of the raw Multivariate Time Series is cumbersome. The algorithms are heavy, slow, and fail to extract all the knowledge available. In that sense, representation techniques reduce the data points but maintain the inner patterns so that the information present in data persists. There exist several approaches to represent Univariate Time Series, but for the multivariate case they are scarce. We have adapted the existent univariate representation algorithm BEATS, which is based on Discrete Cosine transformation and the extraction of eigenvalues to multivariate scenarios and we have called it MultiBEATS. Our method allows flexibility with regard to data reduction and does not require labelled data. To investigate the efficacy of MultiBEATS, we selected 8 open multivariate time series datasets and compared the running time and accuracy of their classification using raw data and MultiBEATS transformed data. We have also compared our results with the representation algorithm Seq2VAR, based on autoencoders, for having a baseline. The MultiBEATS transformation reduces the execution time of classification algorithms, has improved the results in accuracy in 2 of the datasets and matches a third one. With regards to the baseline, MultiBEATS is faster than Seq2VAR and more accurate in 7 out of the 8 datasets. In conclusion, MultiBEATS is a promising approach for efficiently reducing multivariate time series data, therefore helping decision-making in dynamic smart environments.

Anomaly detection in oil-producing wells: a comparative study of one-class classifiers in a multivariate time series dataset

Anomalies in oil-producing wells can have detrimental financial implications, leading to production disruptions and increased maintenance costs. Machine learning techniques offer a promising solution for detecting and preventing such anomalies, minimizing these disruptions and expenses. In this study, we focused on detecting faults in naturally flowing offshore oil and subsea gas-producing wells, utilizing the publicly available 3W dataset comprising multivariate time series data. We conducted a comparison of different anomaly detection methods, specifically one-class classifiers, including Isolation Forest, One-class Support Vector Machine (OCSVM), Local Outlier Factor (LOF), Elliptical Envelope, and Autoencoder with feedforward and LSTM architectures. Our evaluation encompassed two variations: one with feature extraction and the other without, each assessed in both simulated and real data scenarios. Across all scenarios, the LOF classifier consistently outperformed its counterparts. In real instances, the LOF classifier achieved an F1-measure of 87.0% with feature extraction and 85.9% without. In simulated instances, the LOF classifier demonstrated superior performance, attaining F1 measures of 91.5% with feature extraction and 92.0% without. These results show an improvement over the benchmark established by the 3W dataset. Considering the more challenging nature of real data, the inclusion of feature extraction is recommended to improve the effectiveness of anomaly detection in offshore wells. The superior performance of the LOF classifier suggests that the boundaries of normal cases as a single class may be ill-defined, with normal cases better represented by multiple clusters. The statistical analysis conducted further reinforces the reliability and robustness of these findings, instilling confidence in their generalizability to a larger population. The utilization of individual classifiers per instance allows for tailored hyperparameter configurations, accommodating the specific characteristics of each offshore well.

Matrix-based vs. vector-based linear discriminant analysis: A comparison of regularized variants on multivariate time series data

Over the past two decades, matrix-based or bilinear discriminant analysis (BLDA) methods have received much attention. However, it has been reported that the traditional vector-based regularized LDA (RLDA) is still quite competitive and could outperform BLDA on some benchmark datasets. A central question is whether the superiority of the vector-based RLDA would always hold for general matrix data, or is there any type of matrix data on which BLDA would perform better than RLDA? Actually, the reported comparisons are found to suffer from two limitations: (i) the comparisons are only limited to image data, and (ii) regularized RLDA is compared with non-regularized BLDA. In this paper, we break the two limitations and investigate the central question on another type of matrix data, namely multivariate time series (MTS) data. We propose a new two-parameter regularized BLDA (RBLDA) for MTS data classification. To choose the two parameters, we develop an efficient model selection algorithm. The newly proposed RBLDA enables us to perform a fair comparison between vector-based RLDA and matrix-based RBLDA. Experiments on a number of real MTS data sets are conducted to compare RBLDA with RLDA and evaluate the proposed algorithm. The results reveal that the superiority of the vector-based RLDA does not always hold for general matrix data, and RBLDA outperforms RLDA on MTS data. Moreover, the proposed model selection algorithm is efficient, and RBLDA can produce better visualization of MTS data than RLDA.

CO2 concentration forecasting in smart cities using a hybrid ARIMA–TFT model on multivariate time series IoT data

Carbon Dioxide (CO_{2}) is a significant contributor to greenhouse gas emissions and one of the main drivers behind global warming and climate change. In spite of the global economic slowdown due to the COVID-19 pandemic, the global average atmospheric CO_{2} concentration reached a new record high in 2020 with its year-on-year increase being the fifth highest annual increase in 63 years, according to the National Oceanic and Atmospheric Administration. Furthermore, the years 2020 and 2019 were respectively the second and third warmest, while the decade 2010–2019 was the warmest decade ever recorded. In an attempt to curb this climate emergency, many countries and organizations globally have adopted ambitious goals and announced plans to help dramatically reduce CO_{2} emissions. As part of these plans, various innovative smart city projects are being developed, focusing on implementing Internet of Things (IoT) technologies. By collecting sensor-based data, such technologies aim towards automating data-driven decision-making around carbon emission management and reduction. In this work, a hybrid machine learning system, aimed at forecasting CO_2 concentration levels in a smart city environment was developed using a multivariate time series dataset containing IoT sensor measurements of CO_{2}, as well as various environmental factors, taken at every second. The proposed system demonstrated superior performance to similar methods, while also maintaining a high degree of interpretability. More specifically, the approach was empirically compared against other similar approaches in several scenarios and use cases, thus also offering more insight into the predictive capabilities of such state-of-the-art systems. For this comparison, both traditional time series and deep learning approaches were employed, including the current state-of-the-art architectures, such as attention-based, transformer networks. Results demonstrated that, when measured across various settings and metrics, including three different forecasting horizons, the hybrid solution achieved the best overall results, and in some cases, the difference in performance was statistically significant. At the same time, insights from the system’s inner workings were extracted, shedding light on the reasoning behind the model’s predictions and the factors that contribute to them, thus showcasing its transparency. Lastly, throughout the experiments, deep learning approaches illustrated their ability to better handle the multivariate nature of the dataset and in general tended to outperform the traditional time series methods, especially for longer forecasting horizons.

Soft Subspace Based Ensemble Clustering for Multivariate Time Series Data.

Recently, multivariate time series (MTS) clustering has gained lots of attention. However, state-of-the-art algorithms suffer from two major issues. First, few existing studies consider correlations and redundancies between variables of MTS data. Second, since different clusters usually exist in different intrinsic variables, how to efficiently enhance the performance by mining the intrinsic variables of a cluster is challenging work. To deal with these issues, we first propose a variable-weighted K-medoids clustering algorithm (VWKM) based on the importance of a variable for a cluster. In VWKM, the proposed variable weighting scheme could identify the important variables for a cluster, which can also provide knowledge and experience to related experts. Then, a Reverse nearest neighborhood-based density Peaks approach (RP) is proposed to handle the problem of initialization sensitivity of VWKM. Next, based on VWKM and the density peaks approach, an ensemble Clustering framework (SSEC) is advanced to further enhance the clustering performance. Experimental results on ten MTS datasets show that our method works well on MTS datasets and outperforms the state-of-the-art clustering ensemble approaches.

EGNN: Energy-efficient anomaly detection for IoT multivariate time series data using graph neural network

Anomaly detection has been widely applied in Internet of Things (IoT) to guarantee the health of IoT applications. Current studies on anomaly detection focus mainly on measurement design and discovery methods on the cloud, which, however, are associated with issues of computational heaviness and capacity limitation when applied at the network edge. Thus, it becomes important to ensure that the detection is not only accurate but also energy-efficient. To fill this gap, this paper proposes an accurate and Energy-efficient Graph Neural Network based anomaly detection method (EGNN) for IoT multivariate time series data. Specifically, correlations between sensory data upon different IoT devices, which are rarely considered in the literature, are explored through a developed Subgraph Generation Algorithm (SGA) based on graph structure learning. As a result, a dependency graph with multiple subgraphs and their corresponding centres is generated. Thereafter, to reduce anomaly-irrelevant sensory data transmitted in the network, only sensory data upon subgraph centres are utilized for anomaly detection by a computational-light approach, i.e., a multi-layer perceptron based forecasting method. Once an anomaly is detected, sensory data of whole subgraph data are adopted for obtaining accurate anomaly results, by a graph attention based forecasting method. This GNN-based anomaly detection strategy with Mode Switching (GMS) can greatly reduce anomaly-irrelevant data transmission, especially when anomalies occur infrequently. To validate the effectiveness of our mechanism, extensive experiments are conducted upon real-world IoT multivariate time series datasets, and comparison results demonstrate that our technique outperforms the state-of-the-art counterparts in terms of accuracy and energy-efficiency.

Improving position encoding of transformers for multivariate time series classification

Transformers have demonstrated outstanding performance in many applications of deep learning. When applied to time series data, transformers require effective position encoding to capture the ordering of the time series data. The efficacy of position encoding in time series analysis is not well-studied and remains controversial, e.g., whether it is better to inject absolute position encoding or relative position encoding, or a combination of them. In order to clarify this, we first review existing absolute and relative position encoding methods when applied in time series classification. We then proposed a new absolute position encoding method dedicated to time series data called time Absolute Position Encoding (tAPE). Our new method incorporates the series length and input embedding dimension in absolute position encoding. Additionally, we propose computationally Efficient implementation of Relative Position Encoding (eRPE) to improve generalisability for time series. We then propose a novel multivariate time series classification model combining tAPE/eRPE and convolution-based input encoding named ConvTran to improve the position and data embedding of time series data. The proposed absolute and relative position encoding methods are simple and efficient. They can be easily integrated into transformer blocks and used for downstream tasks such as forecasting, extrinsic regression, and anomaly detection. Extensive experiments on 32 multivariate time-series datasets show that our model is significantly more accurate than state-of-the-art convolution and transformer-based models. Code and models are open-sourced at https://github.com/Navidfoumani/ConvTran.

Selective Imputation for Multivariate Time Series Datasets With Missing Values

Multivariate time series often contain missing values for reasons such as failures in data collection mechanisms. Since these missing values can complicate the analysis of time series data, imputation techniques are typically used to deal with this issue. However, the quality of the imputation directly affects the performance of downstream tasks. In this paper, we propose a selective imputation method that identifies a subset of timesteps with missing values to impute in a multivariate time series dataset. This selection, which will result in shorter and simpler time series, is based on both reducing the uncertainty of the imputations and representing the original time series as good as possible. In particular, the method uses multi-objective optimization techniques to select the optimal set of points, and in this selection process, we leverage the beneficial properties of the Multi-task Gaussian Process (MGP). The method is applied to different datasets to analyze the quality of the imputations and the performance obtained in downstream tasks, such as classification or anomaly detection. The results show that much shorter and simpler time series are able to maintain or even improve both the quality of the imputations and the performance of the downstream tasks.

Correlating the Ambient Conditions and Performance Indicators of the LoRaWAN via Surrogate Gaussian Process-Based Bidirectional LSTM Stacked Autoencoder

LoRa’s biggest advantage is its flexibility, which is the ability to increase or decrease data rate and range while decreasing or increasing sensitivity. Whenever propagation conditions change frequently, this function allows the spreading factor to be modified accordingly. Despite their efficiency and scalability, adaptive data rate algorithms ignore and fail to factor in the complex correlation between ambient weather parameters influencing the communication channel desig n. In this research, a Bayesian surrogate Gaussian pr ocess based bidirectional LSTM stacked autoencoder model (BSGP BLSTMSAE) is proposed to estimate the channel perfor-mance indicators such as received signal strength indicator (RSSI) and signal to noise ratio (SNR) and to determine the correlation between the ambient weather conditions and per-formance indicators for the LoRaWAN network. Bayesian op-timization algorithm has been used to optimize the hyper parameters of the developed model. A LoRaWAN experimental multivariate time series dataset has been used for the evalua-tion of the developed model, which upon testing and validation produces high accuracy in predicting the channel performance indicators and ambient conditions of the experimental Lo-RaWAN network. The mean absolu te error of the developed model was around 0.45. Thus, the proposed model can predict the link performance indicators and thereby assist in real time optimization of the transmission parameters to enhance the network performance in LoRaWAN based systems at different ambient conditions