Running Machine Learning Research Articles

An adaptive binary classifier for highly imbalanced datasets on the Edge

Edge machine learning brings intelligence to low-power devices at the periphery of a network. By running machine learning algorithms on the Edge, classification can be performed faster without the need to transmit large data volumes across a network. However, on-device training is often not feasible since Edge devices have limited computing and storage resources. Improved, Scalable, Efficient, and Fast classifieR (iSEFR) is a classifier that performs both training and testing on low-power devices using linearly separable balanced datasets. The novelty of this work is the improvement of the iSEFR accuracy by fine-tuning the algorithm with datasets having an uneven class distribution. Three adaptive linear function transformation techniques were proposed to improve the decision threshold which is in the form of a linear function. Experiments using stratified sampling with 5-fold cross-validation demonstrate that one of the proposed techniques significantly improved F1-score, Recall and Matthews Correlation Coefficient (MCC) by an average of 23 %, 35 % and 21 % compared to iSEFR. Further evaluation of this technique in a Fog environment using highly imbalanced datasets such as credit card fraud, network intrusion and diabetic retinopathy also showed a significant increase of 38 %, 44 % and 30 % in F1-score, Recall and MCC with a Precision of 97 %. The adaptive binary classifier maintained the time complexity of iSEFR without altering the class imbalance.

MicroGravity Explorer Kit (MGX): An Open-Source Platform for Accessible Space Science Experiments

The study of microgravity, a condition in which an object experiences near-zero weight, is a critical area of research with far-reaching implications for various scientific disciplines. Microgravity allows scientists to investigate fundamental physical phenomena influenced by Earth’s gravitational forces, opening up new possibilities in fields such as materials science, fluid dynamics, and biology. However, the complexity and cost of developing and conducting microgravity missions have historically limited the field to well-funded space agencies, universities with dedicated government funding, and large research institutions, creating a significant barrier to entry. This paper presents the MicroGravity Explorer Kit’s (MGX) design, a multifunctional platform for conducting microgravity experiments aboard suborbital rocket flights. The MGX aims to democratize access to microgravity research, making it accessible to high school students, undergraduates, and researchers. To ensure that the tool is versatile across different scenarios, the authors conducted a comprehensive literature review on microgravity experiments, and specific requirements for the MGX were established. The MGX is designed as an open-source platform that supports various experiments, reducing costs and accelerating development. The multipurpose experiment consists of a Jetson Nano computer with multiple sensors, such as inertial sensors, temperature and pressure, and two cameras with up to 4k resolution. The project also presents examples of codes for data acquisition and compression and the ability to process images and run machine learning algorithms to interpret results. The MGX seeks to promote greater participation and innovation in space sciences by simplifying the process and reducing barriers to entry. The design of a platform that can democratize access to space and research related to space sciences has the potential to lead to groundbreaking discoveries and advancements in materials science, fluid dynamics, and biology, with significant practical applications such as more efficient propulsion systems and novel materials with unique properties.

Integration of metagenomics and metaproteomics in the intestinal lavage fluids benefits construction of discriminative model and discovery of biomarkers for HBV liver diseases.

Intestinal lavage fluid (IVF) containing the mucosa-associated microbiota instead of fecal samples was used to study the gut microbiota using different omics approaches. Focusing on the 63 IVF samples collected from healthy and hepatitis B virus-liver disease (HBV-LD), a question is prompted whether omics features could be extracted to distinguish these samples. The IVF-related microbiota derived from the omics data was classified into two enterotype sets, whereas the genomics-based enterotypes were poorly overlapped with the proteomics-based one in either distribution of microbiota or of IVFs. There is lack of molecular features in these enterotypes to specifically recognize healthy or HBV-LD. Running machine learning against the omics data sought the appropriate models to discriminate the healthy and HBV-LD IVFs based on selected genes or proteins. Although a single omics dataset is basically workable in such discrimination, integration of the two datasets enhances discrimination efficiency. The protein features with higher frequencies in the models are further compared between healthy and HBV-LD based on their abundance, bringing about three potential protein biomarkers. This study highlights that integration of metaomics data is beneficial for a molecular discriminator of healthy and HBV-LD, and reveals the IVF samples are valuable for microbiome in a small cohort.

Optimizing Convolutional Neural Networks for Image Classification on Resource-Constrained Microcontroller Units

Running machine learning algorithms for image classification locally on small, cheap, and low-power microcontroller units (MCUs) has advantages in terms of bandwidth, inference time, energy, reliability, and privacy for different applications. Therefore, TinyML focuses on deploying neural networks on MCUs with random access memory sizes between 2 KB and 512 KB and read-only memory storage capacities between 32 KB and 2 MB. Models designed for high-end devices are usually ported to MCUs using model scaling factors provided by the model architecture’s designers. However, our analysis shows that this naive approach of substantially scaling down convolutional neural networks (CNNs) for image classification using such default scaling factors results in suboptimal performance. Consequently, in this paper we present a systematic strategy for efficiently scaling down CNN model architectures to run on MCUs. Moreover, we present our CNN Analyzer, a dashboard-based tool for determining optimal CNN model architecture scaling factors for the downscaling strategy by gaining layer-wise insights into the model architecture scaling factors that drive model size, peak memory, and inference time. Using our strategy, we were able to introduce additional new model architecture scaling factors for MobileNet v1, MobileNet v2, MobileNet v3, and ShuffleNet v2 and to optimize these model architectures. Our best model variation outperforms the MobileNet v1 version provided in the MLPerf Tiny Benchmark on the Visual Wake Words image classification task, reducing the model size by 20.5% while increasing the accuracy by 4.0%.

Heart Disease Prediction Using ML

Machine Learning and artificial intelligence have found valuable on variety of disciplines during their growth, particularly in the light of massive increase in data in recent years. It has the potential to be more dependable in terms of producing quicker and more accurate illness prediction judgments. Therefore, the use of machine learning algorithms to forecast different diseases is growing. Building a model can also aid in the visualization and analysis of diseases to increase the accuracy and consistency of reporting. This article has looked into using several machine learning algorithms to identify cardiac disease. This article's study has demonstrated a step procedure. In a dataset on heart disease initially prepared in the format needed to run machine learning algorithms. The UCI is the source of patient medical records and other data. The presence are absence of heart disease in patients is then ascertained using the heart disease dataset. Second, this paper presents a number of noteworthy findings. The confusion matrix is used to validate the accuracy rate of machine learning methods, including Gradient Boosting Classifier, Support Vector Machine, and Logistic Regression. According to recent research, the Logistic Regression method outperforms other algorithms in terms of accuracy, yielding a high 95% rate. It also outperforms the other four algorithms in terms of recall, precision, and f1-score correctness. The difficult and future research component of this project will be raising the accuracy rates of the machine learning algorithms to between 97% and 100%.

Reducing the Power Consumption of Edge Devices Supporting Ambient Intelligence Applications

Having as a main objective the exploration of power efficiency of microcontrollers running machine learning models, this manuscript contrasts the performance of two types of state-of-the-art microcontrollers, namely ESP32 with an LX6 core and ESP32-S3 with an LX7 core, focusing on the impact of process acceleration technologies like cache memory and vectoring. The research employs experimental methods, where identical machine learning models are run on both microcontrollers under varying conditions, with particular attention to cache optimization and vector instruction utilization. Results indicate a notable difference in power efficiency between the two microcontrollers, directly linked to their respective process acceleration capabilities. The study concludes that while both microcontrollers show efficacy in running machine learning models, ESP32-S3 with an LX7 core demonstrates superior power efficiency, attributable to its advanced vector instruction set and optimized cache memory usage. These findings provide valuable insights for the design of power-efficient embedded systems supporting machine learning for a variety of applications, including IoT and wearable devices, ambient intelligence, and edge computing and pave the way for future research in optimizing machine learning models for low-power, embedded environments.

IMPLEMENTASI ALGORITMA CNN MOBILENET UNTUK KLASIFIKASI GAMBAR SAMPAH DI BANK SAMPAH

Waste is a global problem that must be resolved so that the environment is well maintained, especially in developing countries such as Indonesia. A good understanding of waste will have an impact on optimal waste management. Waste banks are one way out in waste management so that what was originally useless becomes marketable. In its implementation, the waste bank must have data collection and transparency of customer waste data in real time. Technology is one of the main solutions to produce digital products that make it easier for people to access information accurately. By using the Convolutional Neural Network (CNN) algorithm, an image of image data can be predicted with good accuracy. Along with the development of technology today, various kinds of architectures are present, one of which is Mobilenet. This architecture has the ability to run machine learning models on mobile and IoT devices. Furthermore, the resulting model is quite good with an accuracy rate of 96% on Metal waste, 92% on Paper and Organic waste, 80% on Cardboard waste, 76% on Glass waste, and 72% on Plastic waste. The disadvantages that exist in this model when predicting with almost similar shapes and images that have many objects in them, the error will be greater so there is a possibility of error in predicting the results of the garbage image.

An Ultra-Low Power Wearable BMI System with Continual Learning Capabilities.

Driven by the progress in efficient embedded processing, there is an accelerating trend toward running machine learning models directly on wearable Brain-Machine Interfaces (BMIs) to improve portability and privacy and maximize battery life. However, achieving low latency and high classification performance remains challenging due to the inherent variability of electroencephalographic (EEG) signals across sessions and the limited onboard resources. This work proposes a comprehensive BMI workflow based on a CNN-based Continual Learning (CL) framework, allowing the system to adapt to inter-session changes. The workflow is deployed on a wearable, parallel ultra-low power BMI platform (BioGAP). Our results based on two in-house datasets, Dataset A and Dataset B, show that the CL workflow improves average accuracy by up to 30.36% and 10.17%, respectively. Furthermore, when implementing the continual learning on a Parallel Ultra-Low Power (PULP) microcontroller (GAP9), it achieves an energy consumption as low as 0.45mJ per inference and an adaptation time of only 21.5 ms, yielding around 25 h of battery life with a small 100 mAh, 3.7 V battery on BioGAP. Our setup, coupled with the compact CNN model and on-device CL capabilities, meets users' needs for improved privacy, reduced latency, and enhanced inter-session performance, offering good promise for smart embedded real-world BMIs.

Dirençli Punta Kaynağı Prosesinin KNN ve CART Makine Öğrenimi Teknikleri ile Değerlendirilmesi

Spot welding, a type of resistance welding, is a welding application widely used in the production area and it is a common method for joining metal sheets. The spot-welding process is widely used in many production areas, especially in the automotive industry, radiator, and wire mesh production. Spot welding in car production lines is mainly performed by robotic applications. Industry 4.0 and digital transformation trends have led to unprecedented data growth. Nowadays, the manufacturing industry benefits from the power of machine learning and data science algorithms to monitor production processes and make predictions for quality, maintenance, and production optimization. Applying machine learning algorithms reduces the duration and cost of experiments. This study aims to confirm whether the spot welding, applied by robotic arms, is within the ideal spot-welding norms, in real production area. The ideal parameter norms were evaluated by using KNN and CART machine learning algorithms. To use real production data, this study was executed in the body production assembly line, which is selected as the pilot area, at TOFAŞ factory. The data set used in this research consists of the welding parameters of the current year, 2023. By running machine learning algorithms on the dataset, the performance evaluation of each algorithm was examined and the most appropriate estimation method was determined. In the experiments, the best F1-Score value was obtained by the CART model with 93%.

PePTM: An Efficient and Accurate Personalized P2P Learning Algorithm for Home Thermal Modeling

Nowadays, the integration of home automation systems with smart thermostats is a common trend, designed to enhance resident comfort and conserve energy. The introduction of smart thermostats that can run machine learning algorithms has opened the door for on-device training, enabling customized thermal experiences in homes. However, leveraging the flexibility offered by on-device learning has been hindered by the absence of a tailored learning scheme that allows for accurate on-device training of thermal models. Traditional centralized learning (CL) and federated learning (FL) schemes rely on a central server that controls the learning experience, compromising the home’s privacy and requiring significant energy to operate. To address these challenges, we propose PePTM, a personalized peer-to-peer thermal modeling algorithm that generates tailored thermal models for each home, offering a controlled learning experience with a minimal training energy footprint while preserving the home’s privacy, an aspect difficult to achieve in both CL and FL. PePTM consists of local and collaborative learning phases that enable each home to train its thermal model and collaboratively improve it with a set of similar homes in a peer-to-peer fashion. To showcase the effectiveness of PePTM, we use a year’s worth of data from US homes to train thermal models using the RNN time-series model and compare the data across three learning schemes: CL, FL, and PePTM, in terms of model performance and the training energy footprint. Our experimental results show that PePTM is significantly energy-efficient, requiring 695 and 40 times less training energy than CL and FL, respectively, while maintaining comparable performance. We believe that PePTM sets the stage for new avenues for on-device thermal model training, providing a personalized thermal experience with reduced energy consumption and enhanced privacy.

Automatic Gate Opening System with Face Recognition

Despite the limitations of current technology, facial recognition from still or moving sources is developing into a useful tool for law enforcement, security, and counterterrorism applications. Currently, facial recognition shows promise and has been used in a small number of applications, The effectiveness of a facial recognition algorithm is examined in this assessment, specifically with regards to factors such as age, gender, ethnicity, facial characteristics, and lighting levels. Face recognition is being used more frequently as a non-intrusive identity verification technique. In this paper, we propose for the first time a strongly privacy-enhanced face recognition system that effectively uses secure multiparty computation techniques to conceal both the biometrics and the result from the server performing the matching operation. These issues can be easily solved by Explainable AI (XAI), and the solutions are clear to the end users. The proposed system is entirely implemented on a Raspberry Pi, enabling a full embedded application. With the aid of open-source libraries for training, defining, and running machine learning models like Tensorflow.js, Keras, and OpenCV, the application was created using Python and HTML on PyCharm/Visual Studio Code. An entire application can be run on a microcontroller, like the Raspberry Pi, which enables plug-and-play operation of the system at any time. The face detection subsystem, which is based on an improved PCA algorithm, can identify faces in moving vehicles and communicate with the vehicle's owner via GSM. The car's location is sent via GPS. This is the most reliable car security system because we can still locate the car even when it is lost and we can see the thief's face. The owner of the car is free to give the password to new drivers as well. This makes the car security system safer and more comfortable.

Constrained IoT-Based Machine Learning for Accurate Glycemia Forecasting in Type 1 Diabetes Patients.

Individuals with diabetes mellitus type 1 (DM1) tend to check their blood sugar levels multiple times daily and utilize this information to predict their future glycemic levels. Based on these predictions, patients decide on the best approach to regulate their glucose levels with considerations such as insulin dosage and other related factors. Nevertheless, modern developments in Internet of Things (IoT) technology and innovative biomedical sensors have enabled the constant gathering of glucose level data using continuous glucose monitoring (CGM) in addition to other biomedical signals. With the use of machine learning (ML) algorithms, glycemic level patterns can be modeled, enabling accurate forecasting of this variable. Constrained devices have limited computational power, making it challenging to run complex machine learning algorithms directly on these devices. However, by leveraging edge computing, using lightweight machine learning algorithms, and performing preprocessing and feature extraction, it is possible to run machine learning algorithms on constrained devices despite these limitations. In this paper we test the burdens of some constrained IoT devices, probing that it is feasible to locally predict glycemia using a smartphone, up to 45 min in advance and with acceptable accuracy using random forest.

Identifying Chinese social media users' need for affect from their online behaviors.

The need for affect (NFA), which refers to the motivation to approach or avoid emotion-inducing situations, is a valuable indicator of mental health monitoring and intervention, as well as many other applications. Traditionally, NFA has been measured using self-reports, which is not applicable in today's online scenarios due to its shortcomings in fast, large-scale assessments. This study proposed an automatic and non-invasive method for recognizing NFA based on social media behavioral data. The NFA questionnaire scores of 934 participants and their social media data were acquired. Then we run machine learning algorithms to train predictive models, which can be used to automatically identify NFA degrees of online users. The results showed that Extreme Gradient Boosting (XGB) performed best among several algorithms. The Pearson correlation coefficients between predicted scores and NFA questionnaire scores achieved 0.25 (NFA avoidance), 0.31 (NFA approach) and 0.34 (NFA total), and the split-half reliabilities were 0.66-0.70. Our research demonstrated that adolescents' NFA can be identified based on their social media behaviors, and opened a novel way of non-intrusively perceiving users' NFA which can be used for mental health monitoring and other situations that require large-scale NFA measurements.

System for Semi-Automated Literature Review Based on Machine Learning

This paper presents the design and implementation of a system for semi-automating the literature review process based on machine learning. By using machine learning algorithms, the system determines whether scientific papers belong to the topic that is being explored as part of the review process. The system’s user interface allows the process of creating a literature review to be managed through a series of steps: selecting data sources, building queries and topic searches, displaying the scientific papers found, selecting papers that belong to the set of desired papers, running machine learning algorithms for learning and automated classification, and displaying and exporting the final set of papers. Manual literature reviews are compared with automated reviews, and similarities and differences between the two approaches in terms of duration, accuracy, and ease of use are discussed. This study concludes that the best results in terms of sensitivity and accuracy for the automated literature review process are achieved by using a combined machine learning model, which uses multiple unweighted machine learning models. Cross-testing the models on two alternative datasets revealed an overlap in the machine learning hyperparameters. The stable sensitivity and accuracy in the tests indicate the potential for generalized use in automated literature review.

In preprints: improving and interrogating embryo models.

In preprints: improving and interrogating embryo models.

A Novel Plug-in Board for Remote Insect Monitoring

The conventional approach to monitoring insect swarming is based on traps that are periodically checked by human operators. However, human checking of trap contents is expensive, and in many environments, the pest species most frequently encountered in the traps can be detected and monitored automatically. To achieve this goal, a dedicated data acquisition device is necessary, which makes real-time and online pest monitoring possible from a distant location. In addition, it is beneficial for the device to run machine learning algorithms that count and identify insects automatically from pictures. Thanks to the advantages of integrated circuits, more systems have been designed to improve integrated pest management in the context of precision agriculture. However, in our opinion, all of those systems have one or more disadvantages, such as high cost, low power autonomy, low picture quality, a WIFI coverage requirement, intensive human control, and poor software support. Therefore, the aim of this work is to present a novel plug-in board for automatic pest detection and counting. The plug-in board is dedicated to Raspberry Pi devices, especially the Raspberry Pi Zero. The proposed board, in combination with a Raspberry Pi device and a Pi camera, overcomes the limitations of other prototypes found in the literature. In this paper, a detailed description can be found about the schematic and characteristics of the board with a deep-learning-based insect-counting method.

Tiny Machine Learning for Resource-Constrained Microcontrollers

We use 250 billion microcontrollers daily in electronic devices that are capable of running machine learning models inside them. Unfortunately, most of these microcontrollers are highly constrained in terms of computational resources, such as memory usage or clock speed. These are exactly the same resources that play a key role in teaching and running a machine learning model with a basic computer. However, in a microcontroller environment, constrained resources make a critical difference. Therefore, a new paradigm known as tiny machine learning had to be created to meet the constrained requirements of the embedded devices. In this review, we discuss the resource optimization challenges of tiny machine learning and different methods, such as quantization, pruning, and clustering, that can be used to overcome these resource difficulties. Furthermore, we summarize the present state of tiny machine learning frameworks, libraries, development environments, and tools. The benchmarking of tiny machine learning devices is another thing to be concerned about; these same constraints of the microcontrollers and diversity of hardware and software turn to benchmark challenges that must be resolved before it is possible to measure performance differences reliably between embedded devices. We also discuss emerging techniques and approaches to boost and expand the tiny machine learning process and improve data privacy and security. In the end, we form a conclusion about tiny machine learning and its future development.

High-resolution soil moisture mapping using PSO-based optimized cerebellar model articulation controller (CMAC)

A good knowledge in eco-hydrological processes requires significant understanding of geospatial distribution of soil moisture (SM). However, SM monitoring remains challenging due to its large spatial variability and its dynamic time response. This study was performed to assess the performance of a particle swarm optimization (PSO)-based optimized Cerebellar Model Articulation Controller (CMAC) in generating high-resolution surface SM estimates using sentinel-2 imagery over a Mediterranean agro-ecosystem. Furthermore, the results were compared with those of PSO-based optimized group method of data handling (GMDH) as a more common data-driven method. Two different modeling approaches i.e. modeling in homogenous clusters (local approach) and modeling in entire area as an entity (global approach) were examined. Candidate predictors namely sentinel-2 spectral bands, normalized difference vegetation index (NDVI) and normalized difference water index (NDWI), digital elevation model (DEM), slope and aspect were used as the input variables to estimate SM. An intensive field survey had been done to gather in-situ SM data using a time-domain reflectometer (TDR). K-fold validation based on in-situ SM measurements demonstrated the reasonability of the SM estimation of the proposed methodology. Detecting homogeneous areas was done using genetic and particle swarm optimization algorithms. Synthesized SM product of PSO-GMDH showed a mean Normalized Root-Mean-Square Error (NRMSE) of 13.6 to 8.91 for global and local approaches in the test phase. PSO-CMAC method with an average NRMSE of 12.47 to 8.72 for global and local approaches shows the highest accuracy and outperforms the PSO-GMDH method at both local and global approaches. Overall, results revealed that clustering study area prior to running machine learning (ML) models coupled with optical satellite imagery and geophysical properties, boosts their predictive performance and can lead to more accurate mapping of SM with more heterogeneity. The results also showed that the global approach had a moderate performance in capturing the SM heterogeneity.

Crossing the Bridge From Degeneration to Deformity: When Does Sagittal Correction Impact Outcomes in Adult Spinal Deformity Surgery?

Patients with less severe adult spinal deformity (ASD) undergo surgical correction and often achieve good clinical outcomes. However, it is not well understood how much clinical improvement is due to sagittal correction rather than treatment of the spondylotic process. Determine baseline thresholds in radiographic parameters that, when exceeded, may result in substantive clinical improvement from surgical correction. Retrospective. ASD patients with BL and two-year data were included. Parameters assessed: sagittal vertical axis, pelvic incidence-lumbar lordosis mismatch, pelvic tilt, T1 pelvic angle, L1 pelvic angle, L4-S1 lordosis, C2-C7 sagittal vertical axis, C2-T3, C2 slope. Outcomes: Good Outcome (GO) at two years: [meeting either: (1) Substantial Clinical Benefit for Oswestry Disability Index (change >18.8), or (2) Oswestry Disability Index <15 and Scoliosis Research Society Total>4.5]. Binary logistic regression assessed each parameter to determine if correction was more likely needed to achieve GO. Conditional inference tree run machine learning analysis generated baseline thresholds for each parameter, above which, correction was necessary to achieve GO. We included 431 ASD patients. There were 223 (50%) that achieved a GO by two years. Binary logistic regression analysis demonstrated, with increasing baseline severity in deformity, sagittal correction was more often seen in those achieving GO for each parameter(all P <0.001). Of patients with baseline T1 pelvic angle above the threshold, 95% required correction to meet GO (95% vs. 54%, P <0.001). A baseline pelvic incidence-lumbar lordosis >10° (74% of patients meeting GO) needed correction to achieve GO (odds ratio: 2.6, 95% confidence interval: 1.4-4.8). A baseline C2 slope >15° also necessitated correction to obtain clinical success (odds ratio: 7.7, 95% confidence interval: 3.7-15.7). Our study highlighted point may be present at which sagittal correction has an outsized influence on clinical improvement, reflecting the line where deformity becomes a significant contributor to disability. These new thresholds give us insight into which patients may be more suitable for sagittal correction, as opposed to intervention for the spondylotic process only, leading to a more efficient utility of surgical intervention for ASD.

P113. At what point does degenerative become deformity: when good outcomes necessitate sagittal correction in adult cervical deformity surgery

BACKGROUND CONTEXT Current radiographic thresholds have shown inconsistent correlation with clinical outcomes. However, it has not been determined at what point in presenting deformity does sagittal correction tremendously improve clinical outcomes following cervical deformity surgery. PURPOSE Define baseline thresholds in cervical parameters that, when exceeded, benefit dramatically from surgical correction to achieve ideal clinical outcomes. STUDY DESIGN/SETTING Retrospective. PATIENT SAMPLE A total of 290 ACD patients. OUTCOME MEASURES Clinical outcomes, radiographic alignment. METHODS CD patients with baseline (BL) and 2-year (2Y) data included. Parameters assessed: C2 slope (C2S), C2-C7 Lordosis, C2-C7 SVA (cSVA), T1 Slope (T1S), TS-CL. Outcomes: Virket al Good Clinical Outcome (GCO): [Meeting 2 of 3: 1) an NDI>20 or meeting MCID, 2) mJOA >=14), 3) an NRS-Neck<=5 or improved by 2 or more points from baseline]. Binary logistic regression assessed each parameter to determine if correction was more likely needed to achieve GCO. Conditional inference tree (CIT) run machine learning analysis generated baseline thresholds for each parameter, above which, correction was necessary to achieve GCO. RESULTS Included: 105 CD patients. There were 57 (54%) of patients achieving GCO by 2Y. Correction was necessitated when baseline C2S was above 20° (OR: 6.8, [1.6-28.9]; p=.01) and when baseline C2-C7 Lordosis was below 10° (OR: 16, [2.4-107.5]; p=.004). Patients presenting with a cSVA above 20 mm more often achieved clinical success when corrected (74.2% vs 0.0%, p<.001). A baseline T1 slope above 23° was 16 times more likely to reach GCO with correction than those below this threshold (p=.005). TS-CL more likely required correction to reach GCO when above 26° at baseline (OR: 7.0, [1.7-29.1]; p=.007). When assessing patients above both the cSVA and C2S threshold versus the remaining cohort, these patients more likely met GCO when corrected in either parameter (OR: 22.5, [3.3-152.0]; p=.001). CONCLUSIONS Our study highlighted the importance of correction and the threshold at which it dramatically impacts clinical success. These new thresholds delineate patients obtaining superior benefit for sagittal correction and may better increase the utility gained from surgical intervention for cervical deformity. FDA DEVICE/DRUG STATUS This abstract does not discuss or include any applicable devices or drugs. Current radiographic thresholds have shown inconsistent correlation with clinical outcomes. However, it has not been determined at what point in presenting deformity does sagittal correction tremendously improve clinical outcomes following cervical deformity surgery. Define baseline thresholds in cervical parameters that, when exceeded, benefit dramatically from surgical correction to achieve ideal clinical outcomes. Retrospective. A total of 290 ACD patients. Clinical outcomes, radiographic alignment. CD patients with baseline (BL) and 2-year (2Y) data included. Parameters assessed: C2 slope (C2S), C2-C7 Lordosis, C2-C7 SVA (cSVA), T1 Slope (T1S), TS-CL. Outcomes: Virket al Good Clinical Outcome (GCO): [Meeting 2 of 3: 1) an NDI>20 or meeting MCID, 2) mJOA >=14), 3) an NRS-Neck<=5 or improved by 2 or more points from baseline]. Binary logistic regression assessed each parameter to determine if correction was more likely needed to achieve GCO. Conditional inference tree (CIT) run machine learning analysis generated baseline thresholds for each parameter, above which, correction was necessary to achieve GCO. Included: 105 CD patients. There were 57 (54%) of patients achieving GCO by 2Y. Correction was necessitated when baseline C2S was above 20° (OR: 6.8, [1.6-28.9]; p=.01) and when baseline C2-C7 Lordosis was below 10° (OR: 16, [2.4-107.5]; p=.004). Patients presenting with a cSVA above 20 mm more often achieved clinical success when corrected (74.2% vs 0.0%, p<.001). A baseline T1 slope above 23° was 16 times more likely to reach GCO with correction than those below this threshold (p=.005). TS-CL more likely required correction to reach GCO when above 26° at baseline (OR: 7.0, [1.7-29.1]; p=.007). When assessing patients above both the cSVA and C2S threshold versus the remaining cohort, these patients more likely met GCO when corrected in either parameter (OR: 22.5, [3.3-152.0]; p=.001). Our study highlighted the importance of correction and the threshold at which it dramatically impacts clinical success. These new thresholds delineate patients obtaining superior benefit for sagittal correction and may better increase the utility gained from surgical intervention for cervical deformity.