Description:TECHNICAL FIELD
[0001] The present disclosure relates to the field of machine learning, specifically to a system and a method for enhancing performance of a metamodel to enhance the performance and generalization of predictive models used in classification and regression tasks.

BACKGROUND
[0002] Over past several decades, machine learning techniques have been widely adopted in a variety of applications ranging from image recognition and natural language processing to predictive analytics. Conventional methods in this field typically involve training models on datasets that are partitioned into distinct subsets, commonly designated as training and testing sets, to evaluate model performance and generalization capabilities. Despite significant progress, many established approaches continue to face challenges such as overfitting, where a model performs exceptionally well on the training data yet fails to replicate that performance when exposed to new, unseen data.
[0003] In response to overfitting, conventional methods developed various ensemble techniques that combine multiple predictive models in an effort to improve overall accuracy and robustness. The conventional methods such as stacking and cross-validation have also been utilized to aggregate the strengths of individual base models. Such approaches aim to reduce variance and better capture the underlying statistical patterns by leveraging well-known principles, including a central limit theorem and a law of large numbers, which help in mitigating issues associated with small or heterogeneous datasets.
[0004] However, as the complexity of real-world data increases, characterized by varying sample sizes, high-dimensional feature spaces, and inherent noise, limitations of the conventional methods become more pronounced. Existing techniques may not sufficiently address the variability found in practical scenarios, particularly when datasets are either limited in size or exhibit significant diversity. As a result, there remains a continuing need for improved strategies that can adapt to different data characteristics and enhance the generalization of predictive models without incurring the pitfalls of overfitting.
[0005] To overcome at least the aforementioned limitations, there is a need for a machine learning methodology that eliminates the shortcomings of conventional ensemble techniques.

OBJECTS OF THE PRESENT DISCLOSURE
[0006] A general object of the present disclosure is to provide a method and a system for enhancing performance and generalization of a metamodel used in machine learning applications.
[0007] Another object of the present disclosure is to enable dynamic selection of base classifiers and regressors based on one or more characteristics of the dataset.
[0008] Another object of the present disclosure is to reduce overfitting by employing a shuffling process that selectively redistributes prediction data between training and testing datasets.
[0009] Another object of the present disclosure is to improve predictive accuracy and robustness by combining diverse predictions from multiple base models into a unified metamodel.

SUMMARY
[0010] Aspects of the present disclosure generally relates to the field of machine learning, specifically to a system and a method for enhancing performance of a metamodel to enhance the performance and generalization of predictive models used in classification and regression tasks.
[0011] In an aspect, the present disclosure relates to a method for enhancing performance of a metamodel. The method includes partitioning, by a system, a dataset into a training dataset (trB) and a testing dataset (tsB). The method also includes selecting, by the system, at least one of: a plurality of base classifiers and a plurality of base regressors from a classifier set and a regressor set, respectively, upon partitioning the dataset. The method includes training, by the system, each of the plurality of base classifiers and each of the plurality of base regressors on the training dataset. The method includes evaluating, by the system, each of the plurality of base classifiers and each of the plurality of base regressors on the testing dataset. Further, the method includes collecting, by the system, predictions from each of the plurality of base classifiers and each of the plurality of base regressors trained and evaluated on both the training dataset and the testing dataset. The method includes shuffling, by the system, the predictions to generate metamodel training datasets and metamodel testing datasets. The method includes training, by the system, the metamodel using the metamodel training datasets, and evaluating the metamodel using the metamodel testing datasets. The method includes assessing, by the system, performance of the metamodel, by determining one or more evaluation metrics of the metamodel, to enhance accuracy and generalization of the metamodel.
[0012] In one embodiment, the method may include selecting, by the system, the plurality of base classifiers and the plurality of base regressors from the classifier set and the regressor set dynamically based on one or more characteristics of the dataset.
[0013] In one embodiment, the method includes storing, by the system, the predictions as train data predictions and test data predictions in a database, upon collecting the predictions from each of the plurality of base classifiers and each of the plurality of base regressors.
[0014] In one embodiment, the method may include selectively transferring, by the system, a randomly selected subset of testing data into the training dataset, or a predefined set of training data into the testing dataset, for shuffling the predictions.
[0015] In one embodiment, the method may include enabling, by the system, the metamodel to capture underlying patterns of the dataset by training and evaluating the metamodel using the metamodel training datasets and the metamodel testing datasets, respectively.
[0016] In one embodiment, the method may include assessing, by the system, the performance of the metamodel using a Mean Squared Error (MSE) metric.
[0017] In one embodiment, the one or more evaluation metrics may include at least one of: a precision, a recall, and a F1-score.
[0018] In another aspect, the present disclosure relates to a system for enhancing performance of a metamodel. The system includes a processor and a memory. The memory is operatively coupled with the processor. Further, the memory includes one or more instructions which, when executed, cause the processor to partition a dataset into a training dataset (trB) and a testing dataset (tsB). The processor is configured to, upon partitioning the dataset, select at least one of: a plurality of base classifiers and a plurality of base regressors from a classifier set and a regressor set, respectively. The processor is configured to train each of the plurality of base classifiers and each of the plurality of base regressors on the training dataset. The processor is configured to evaluate each of the plurality of base classifiers and each of the plurality of base regressors on the testing dataset. The processor is configured to collect predictions from each of the plurality of base classifiers and each of the plurality of base regressors trained and evaluated on both the training dataset and the testing dataset. The processor is configured to shuffle the predictions to generate metamodel training datasets and metamodel testing datasets. The processor is configured to train the metamodel using the metamodel training datasets, and evaluate the metamodel using the metamodel testing datasets. The processor is configured to assess performance of the metamodel, by determining one or more evaluation metrics of the metamodel, to enhance accuracy and generalization of the metamodel.
[0019] In one embodiment, the processor may be configured to select the plurality of base classifiers and the plurality of base regressors from the classifier set and the regressor set dynamically based on one or more characteristics of the dataset.
[0020] In one embodiment, the processor is configured to shuffle the predictions by selectively transferring a randomly selected subset of testing data into the training dataset, or a predefined set of training data into the testing dataset.
[0021] Various objects, features, aspects, and advantages of the inventive subject matter will become more apparent from the following detailed description of preferred embodiments, along with the accompanying drawing figures in which like numerals represent like components.

BRIEF DESCRIPTION OF THE DRAWINGS
[0022] The accompanying drawings, which are incorporated herein, and constitute a part of this disclosure, illustrate exemplary embodiments of the disclosed methods and systems which like reference numerals refer to the same parts throughout the different drawings. Components in the drawings are not necessarily to scale, emphasis instead being placed upon clearly illustrating the principles of the present disclosure. Some drawings may indicate the components using block diagrams and may not represent the internal circuitry of each component. It will be appreciated by those skilled in the art that disclosure of such drawings includes the disclosure of electrical components, electronic components, or circuitry commonly used to implement such components.
[0023] FIG. 1 illustrates an exemplary computer system in which or with which embodiments of the present disclosure may be utilized , in accordance with an embodiment of the present disclosure.
[0024] FIG. 2 illustrates an exemplary block diagram of a server system, in accordance with an embodiment of the present disclosure.
[0025] FIG. 3 illustrates a flow diagram illustrating a method for enhancing performance of a metamodel, in accordance with an embodiment of the present disclosure.
[0026] FIG. 4 illustrates a schematic flow diagram depicting a process for enhancing performance of a metamodel, in accordance with an embodiment of the present disclosure.
[0027] FIG. 5A illustrates a schematic representation depicting a pattern distribution indicative of overfitting, in accordance with an embodiment of the present disclosure.
[0028] FIG. 5B illustrates a schematic representation depicting a pattern distribution indicative of a regular fit, in accordance with an embodiment of the present disclosure.
[0029] The foregoing shall be more apparent from the following more detailed description of the disclosure.

DETAILED DESCRIPTION
[0030] In the following description, for the purposes of explanation, various specific details are set forth in order to provide a thorough understanding of embodiments of the present disclosure. It will be apparent, however, that embodiments of the present disclosure may be practiced without these specific details. Several features described hereafter can each be used independently of one another or with any combination of other features. An individual feature may not address all of the problems discussed above or might address only some of the problems discussed above. Some of the problems discussed above might not be fully addressed by any of the features described herein.
[0031] The ensuing description provides exemplary embodiments only and is not intended to limit the scope, applicability, or configuration of the disclosure. Rather, the ensuing description of the exemplary embodiments will provide those skilled in the art with an enabling description for implementing an exemplary embodiment. It should be understood that various changes may be made in the function and arrangement of elements without departing from the spirit and scope of the disclosure as set forth.
[0032] The following is a detailed description of embodiments of the disclosure depicted in the accompanying drawings. The embodiments are in such detail as to clearly communicate the disclosure. If the specification states a component or feature “may”, “can”, “could”, or “might” be included or have a characteristic, that particular component or feature is not required to be included or have the characteristic.
[0033] As used in the description herein and throughout the claims that follow, the meaning of “a,” “an,” and “the” includes plural reference unless the context clearly dictates otherwise. Also, as used in the description herein, the meaning of “in” includes “in” and “on” unless the context clearly dictates otherwise.
[0034] Modern machine learning approaches, particularly those involving ensemble techniques such as stacking and cross-validation, often struggle with issues of overfitting and limited generalization, especially when faced with small or heterogeneous datasets. In conventional systems, models tend to capture patterns specific to the training data, resulting in performance degradation when encountering unseen data. Additionally, static selection processes for base models may fail to adequately adjust to varying dataset characteristics, thereby exacerbating performance inconsistencies and undermining the robustness of predictive outcomes.
[0035] The present disclosure proposes a method and a system that address the above-mentioned challenges by integrating dynamic selection, prediction shuffling, and layered ensemble learning. In one embodiment, the method includes partitioning dataset into training and testing subsets, followed by dynamically selecting a plurality of base classifiers and regressors based on unique characteristics of the dataset. Each base model is independently trained and evaluated, with predictions from both training and testing phases being collected and selectively shuffled to generate new, diversified training and testing sets for a higher-level metamodel. The proposed method and system provides an adaptive approach that enables the metamodel to synthesize a broader spectrum of predictive insights, thereby mitigating overfitting and more accurately capturing the underlying data patterns.
[0036] The integrated method and system offer a significant technical advantage by markedly enhancing the generalization capabilities of predictive models. By combining dynamic model selection with a shuffling process, the approach reduces propensity for overfitting while simultaneously boosting predictive accuracy across a diverse range of datasets. Moreover, the system’s design facilitates efficient implementation and scalability, ensuring optimal utilization of computational resources while maintaining robust performance. Such comprehensive framework not only overcomes the deficiencies inherent in conventional ensemble methods but also provides more reliable and adaptable machine learning applications.
[0037] The following is a detailed description of embodiments of the disclosure depicted in the accompanying drawings. The embodiments are in such detail as to clearly communicate the disclosure. However, the amount of detail offered is not intended to limit the anticipated variations of embodiments; on the contrary, the intention is to cover all modifications, equivalents, and alternatives falling within the scope of the present disclosures as defined by the appended claims.
[0038] Embodiments explained herein relate to a system and a method for enhancing performance of a metamodel to enhance the performance and generalization of predictive models used in classification and regression tasks.
[0039] The various embodiments throughout the disclosure will be explained in more detail with reference to FIGs. 1-5.
[0040] Referring to FIG. 1, an exemplary block diagram of a system (100), for enhancing performance of a metamodel is illustrated, in accordance with one or more embodiments of the present disclosure. The system (100) may include a memory (130) (may also be referred to as a main memory (130)) and a processor (170). It should be noted that the system (100) may be a computer system (as per FIG. 1). Further, the system (100) may also be referred to as a server system (200) (as shown in FIG. 2).
[0041] In one exemplary embodiment, the system (100) may also include an external storage device (110), a bus (120), a read-only memory (140), a mass storage device (150), and a communication port(s) (160). A person skilled in the art will appreciate that the system (100) may include more than one processor and communication ports. The processor (170) may include various modules associated with embodiments of the present disclosure. The communication port(s) (160) may be any of an RS-232 port for use with a modem-based dialup connection, a 10/100 Ethernet port, a Gigabit or 10 Gigabit port using copper or fiber, a serial port, a parallel port, or other existing or future ports. The communication ports(s) (160) may be chosen depending on a network, such as a Local Area Network (LAN), Wide Area Network (WAN), or any network to which the system (100) connects.
[0042] In one exemplary embodiment, the main memory (130) may be a Random Access Memory (RAM), or any other dynamic storage device commonly known in the art. The read-only memory (140) may be any static storage device(s) e.g., but not limited to, a Programmable Read Only Memory (PROM) chip for storing static information e.g., start-up or basic input/output system (BIOS) instructions for the processor (170). The mass storage device (150) may be any current or future mass storage solution, which can be used to store information and/or instructions. Exemplary mass storage solutions include, but are not limited to, Parallel Advanced Technology Attachment (PATA) or Serial Advanced Technology Attachment (SATA) hard disk drives or solid-state drives (internal or external, e.g., having Universal Serial Bus (USB) and/or Firewire interfaces).
[0043] In one exemplary embodiment, the bus (120) may communicatively couple the processor(s) (170) with the other memory, storage, and communication blocks. The bus (120) may be, e.g. a Peripheral Component Interconnect (PCI)/PCI Extended (PCI-X) bus, Small Computer System Interface (SCSI), USB, or the like, for connecting expansion cards, drives, and other subsystems as well as other buses, such a front side bus (FSB), which connects the processor (170) to the system (100).
[0044] In one exemplary embodiment, operator, and administrative interfaces, e.g., a display, keyboard, and cursor control device may also be coupled to the bus (120) to support direct operator interaction with the system (100). Other operator and administrative interfaces can be provided through network connections connected through the communication port(s) (160). Components described above are meant only to exemplify various possibilities. In no way should the aforementioned exemplary system (100) limit the scope of the present disclosure.
[0045] In one embodiment, the memory (130) may include one or more instructions which, when executed, cause the processor (170) to partition a dataset into a training dataset (trB) and a testing dataset (tsB). The processor (170) may also be configured to select at least one of a plurality of base classifiers and a plurality of base regressors from a classifier set and a regressor set, respectively, upon partitioning the dataset. The processor (170) may be configured to train each of the plurality of base classifiers and each of the plurality of base regressors on the training dataset. The processor (170) may be configured to evaluate each of the plurality of base classifiers and each of the plurality of base regressors on the testing dataset. The processor (170) may be configured to collect predictions from each of the plurality of base classifiers and each of the plurality of base regressors trained and evaluated on both the training dataset and the testing dataset. Further, the processor (170) may be configured to shuffle the predictions to generate metamodel training datasets and metamodel testing datasets. The processor (170) may be configured to train the metamodel using the metamodel training datasets, and evaluate the metamodel using the metamodel testing datasets. The processor (170) may be configured to assess performance of the metamodel, by determining one or more evaluation metrics of the metamodel, to enhance accuracy and generalization of the metamodel. The one or more evaluation metrics may include, but not limited to, a precision, a recall, and a F1-score.
[0046] In one embodiment, the processor (170) may be configured to select the plurality of base classifiers and the plurality of base regressors from the classifier set and the regressor set dynamically based on one or more characteristics of the dataset.
[0047] In one embodiment, the processor (170) may be configured to store the predictions as train data predictions and test data predictions in a database, upon collecting the predictions from each of the plurality of base classifiers and each of the plurality of base regressors.
[0048] In one embodiment, the processor (170) may be configured to shuffle the predictions by selectively transferring a randomly selected subset of testing data into the training dataset, or a predefined set of training data into the testing dataset.
[0049] In one embodiment, the processor (170) may be configured to enable the metamodel to capture underlying patterns of the dataset by training and evaluating the metamodel using the metamodel training datasets and the metamodel testing datasets, respectively.
[0050] In one embodiment, the processor (170) may be configured to assess the performance of the metamodel using a Mean Squared Error (MSE) metric.
[0051] In one embodiment, the one or more evaluation metrics may include at least one of: a precision, a recall, and a F1-score.
[0052] Referring to FIG. 2, an exemplary block diagram of a server system (200) is illustrated, in accordance with one or more embodiments of the present disclosure. The server system (200) may include one or more processor(s) (202). The one or more processor(s) (202) may be implemented as one or more microprocessors, microcomputers, microcontrollers, digital signal processors, central processing units, logic circuitries, and/or any devices that process data based on operational instructions. Among other capabilities, the one or more processor(s) (202) may be configured to fetch and execute computer-readable instructions stored in a memory (204) of the server system (102). The memory (204) may be configured to store one or more computer-readable instructions or routines in a non-transitory computer readable storage medium, which may be fetched and executed to create or share data packets over a network service. The memory (204) may include any non-transitory storage device including, for example, volatile memory such as random-access memory (RAM), or non-volatile memory such as erasable programmable read only memory (EPROM), flash memory, and the like
[0053] In an embodiment, the server system (102) may include an interface(s) (206). The interface(s) (206) may include a variety of interfaces, for example, interfaces for data input and output devices (I/O), storage devices, and the like. The interface(s) (206) may facilitate communication through the system (108). The interface(s) (206) may also provide a communication pathway for one or more components of the system (108). Examples of such components include, but are not limited to, processing engine(s) (208) and a database (210).
[0054] In an embodiment, the processing engine(s) (208) may be implemented as a combination of hardware and programming (for example, programmable instructions) to implement one or more functionalities of the processing engine(s) (208). In examples described herein, such combinations of hardware and programming may be implemented in several different ways. For example, the programming for the processing engine(s) (208) may be processor-executable instructions stored on a non-transitory machine-readable storage medium and the hardware for the processing engine(s) (208) may include a processing resource (for example, one or more processors), to execute such instructions. In the present examples, the machine-readable storage medium may store instructions that, when executed by the processing resource, implement the processing engine(s) (208). In such examples, the system may include the machine-readable storage medium storing the instructions and the processing resource to execute the instructions, or the machine-readable storage medium may be separate but accessible to the system and the processing resource. In other examples, the processing engine(s) (208) may be implemented by electronic circuitry.
[0055] A person skilled in the art may understand that one or more steps of the present disclosure performed by the processor (170) may also be performed by the one or more processor(s) of the sever system (200). Herein, the one or more instructions may be stored in the memory (204). When the one or more instructions executed, by the processor (202), causes the processor (202) to perform the one or more steps, performed by the processor (170).
[0056] Referring now to FIG.3, a flow diagram for implementing the steps of a proposed method (300) for enhancing performance of a metamodel is illustrated, in accordance with one or more embodiments of the present disclosure. It should be noted that some or all steps (302 to 316) of the proposed method (300) may be performed by the system (100) (i.e., the computer system (100) or the server system (200)).
[0057] The method (300) may include one or more steps for enhancing performance of the metamodel. The one or more steps may be performed by the system (100) as illustrated in FIG. 1.
[0058] At step (302), the method (300) may include partitioning a dataset into a training dataset (trB) and a testing dataset (tsB).
[0059] At step (304), the method (300) may include upon partitioning (302) the dataset, selecting at least one of a plurality of base classifiers and a plurality of base regressors from a classifier set and a regressor set, respectively. The method (300) may also include selecting the plurality of base classifiers and the plurality of base regressors from the classifier set and the regressor set dynamically based on one or more characteristics of the dataset.
[0060] At step (306), the method (300) may include training each of the plurality of base classifiers or each of the plurality of base regressors on the training dataset.
[0061] At step (308), the method (300) may include evaluating each of the plurality of base classifiers or each of the plurality of base regressors on the testing dataset.
[0062] At step (310), the method (300) may include collecting predictions from each of the plurality of base classifiers or each of the plurality of base regressors trained and evaluated on both the training dataset and the testing dataset. Further, the method (300) may include storing the predictions as train data predictions and test data predictions in a database (for example database (210) as illustrated in FIG. 2), upon collecting (310) the predictions from each of the plurality of base classifiers and each of the plurality of base regressors.
[0063] At step (312), the method (300) may include shuffling the predictions to generate metamodel training datasets and metamodel testing datasets. The method (300) may include selectively transferring a randomly selected subset of testing data into the training dataset, or a predefined set of training data into the testing dataset, for shuffling the predictions.
[0064] At step (314), the method (300) may include training the metamodel using the metamodel training datasets, and evaluating the metamodel using the metamodel testing datasets.
[0065] At step (316), the method (300) may include assessing performance of the metamodel, by determining one or more evaluation metrics of the metamodel, to enhance accuracy and generalization of the metamodel. The method (300) may include assessing the performance of the metamodel using a Mean Squared Error (MSE) metric. The one or more evaluation metrics may include, but not limited to, at least one of a precision, a recall, and a F1-score.
[0066] Further, the method (300) may include enabling the metamodel to capture underlying patterns of the dataset by training and evaluating the metamodel using the metamodel training datasets and the metamodel testing datasets, respectively.
[0067] It should also be noted that the steps for executing the method (300) described herein are not limited to the specific steps outlined above. The method may be implemented in various other ways, and the steps may be reordered, combined, or modified without departing from the scope and spirit of the invention. The examples provided are for illustrative purposes only and are not intended to limit the invention to the specific embodiments disclosed. Those skilled in the art will recognize that various modifications and adaptations can be made to the method without departing from the broader inventive concepts disclosed herein.
[0068] Referring to FIG. 4, a schematic block diagram of a process (400) for enhancing performance of the metamodel is illustrated, in accordance with one or more embodiments of the present disclosure. The process (400) outlines a sequential process for generalization of the metamodel, enhanced for both classification and regression tasks. The process (400) may include partitioning comprehensive dataset, denoted as DBm, into two distinct subsets namely a training dataset (trB) and a testing dataset (tsB). The partitioning step is fundamental to establishing the foundation for subsequent training and evaluation processes. Following partitioning, the system (100) may dynamically select a plurality of base models such as classifiers (e.g., KNN, MLP, NB, RF, DT, LR, SVM, and AB) for classification or corresponding regressors for regression from a predefined model set (may also be referred to as the base model (Bm)). Each selected base model may then be independently trained using the training dataset (trB) and evaluated on the testing dataset (tsB).
[0069] Subsequent to the individual model evaluations, the predictions generated by each base model on both trB and tsB may be systematically collected and stored as distinct prediction datasets i.e., TRAIN_DATA_PREDi and TEST_DATA_PREDi. Herein, “i” may refer to a base model number in sequence. The trB (TRAIN_DATA_PREDi) and tsB (TEST_DATA_PREDi) datasets, representing the outcomes of both the training and testing phases respectively, may then be subjected to a shuffling process. The shuffling step may be configured to selectively redistribute portions of the prediction data between the original training and testing sets, thereby creating new, diversified datasets for metamodel training (referred as DTRAINMETA OR trM) and evaluation (referred as DTESTMETA or tsM). The shuffling step is critical, as to enhance the diversity of the input data to the metamodel, mitigating overfitting by ensuring that the metamodel is exposed to a broader spectrum of prediction patterns and underlying data characteristics.
[0070] In one embodiment, the dataset shuffling process may involve creating new training and test datasets for the metamodel by redistributing data from the base model’s datasets. In one embodiment, the process (400) may involve selectively transferring portions of the base model’s test data into the metamodel’s training set. The amount of data transferred may vary, allowing for different levels of mixing from a small fraction to the entire test set. For the same distribution and with minimal transfers, generalization does not occur. By rearranging the data in such way, the process (400) may introduce variability in the metamodel’s training data, exposing the metamodel to diverse prediction patterns from different models and datasets. One objective of the present disclosure is to improve the metamodel’s ability to generalize and improve overall robustness.
[0071] Finally, the system (100) may employ the newly generated metamodel training dataset (trM) to train a higher-level metamodel, which integrates the diverse predictions from the base models. The performance of the metamodel may subsequently be evaluated using the corresponding testing dataset (tsM), where metrics such as precision, recall, and F1-score may be computed for classification tasks, or the Mean Squared Error (MSE) may be determined for regression tasks. The overall technical effect of this integrated process, as depicted in FIG. 4, is a robust, adaptive process that enhances predictive accuracy and generalization, thereby addressing the limitations associated with conventional ensemble techniques.
[0072] Referring to FIG. 5, an exemplary comparison of distribution patterns is illustrated in accordance with one or more embodiments of the present disclosure. FIG. 5A shows a pattern distribution that indicates overfitting (OF) i.e., a scenario where a classification model exhibits overfitting due to an imbalanced distribution of training samples, characterized by a skewed representation of four classes (A, B, C and D). In FIG. 5A, the diagonal points represent correctly classified training samples, and the number of the diagonal points significantly exceeds the number of off-diagonal misclassifications. A biased distribution, as depicted in FIG. 5A, when provided as input to the metamodel, may result in the metamodel predominantly learning from the overrepresented diagonal data. Such a learning process may increase the risk of overfitting and lead to poorer generalization when the metamodel is applied to new, unseen data.
[0073] In contrast, FIG. 5B illustrates a scenario of regular fit (RF) in which the distribution of training samples is balanced. FIG. 5B reveals that both the diagonal and off-diagonal points are equitably represented across the grid, indicating that the training dataset captures the diversity of the underlying data space with accuracy. A metamodel that is trained on a balanced dataset, as shown in FIG. 5B, may be less prone to overfitting because the metamodel learns from a broader variety of data points. Consequently, the balanced training dataset may enhance the metamodel’s ability to generalize effectively to new data and may improve overall performance compared to training based on a skewed distribution.
Therefore, the metamodel may combine stacking and shuffling techniques to eliminate overfitting and improve generalization over traditional cross-validation methods.
[0074] It will be appreciated that one or more additional components may be incorporated, modified, or omitted in the implementation of the present disclosure without departing from the scope as defined by the appended claims. The described embodiments are merely illustrative, and variations in design, structure, or material selection may be made to suit specific applications. Any such modifications, equivalents, or substitutions are intended to be within the scope and spirit of the present invention as defined by the claims.
[0075] While the foregoing describes various embodiments of the present disclosure, other and further embodiments of the present disclosure may be devised without departing from the basic scope thereof. The scope of the present disclosure is determined by the claims that follow. The present disclosure is not limited to the described embodiments, versions, or examples, which are included to enable a person having ordinary skill in the art to make and use the present disclosure when combined with information and knowledge available to the person having ordinary skill in the art.

ADVANTAGES OF THE PRESENT DISCLOSURE
[0076] The present disclosure provides a metamodeling approach that significantly reduces overfitting, thereby enhancing generalization and predictive accuracy of a metamodel compared to conventional ensemble techniques.
[0077] The present disclosure provides dynamic selection of base classifiers and regressors based on dataset characteristics, resulting in a more adaptable and efficient modeling process.
[0078] The present disclosure provides improved robustness across diverse data scenarios by employing a shuffling process that diversifies training data and mitigates noise-related performance degradation.
[0079] The present disclosure provides enhanced performance through the integration of predictions from multiple base models into a unified metamodel, leading to superior overall accuracy and reliability.
[0080] The present disclosure provides an efficient computational system for implementing the proposed method, reducing training time and optimizing resource utilization for large-scale machine learning applications.
, Claims:1. A method (300) for enhancing performance of a metamodel, the method (300) comprising:
partitioning (302), by a system (100), a dataset into a training dataset (trB) and a testing dataset (tsB);
upon partitioning the dataset, selecting (304), by the system (100), at least one of: a plurality of base classifiers and a plurality of base regressors from a classifier set and a regressor set, respectively;
training (306), by the system (100), each of the plurality of base classifiers or each of the plurality of base regressors on the training dataset;
evaluating (308), by the system (100), each of the plurality of base classifiers or each of the plurality of base regressors on the testing dataset;
collecting (310), by the system (100), predictions from each of the plurality of base classifiers or each of the plurality of base regressors trained and evaluated on both the training dataset and the testing dataset;
shuffling (312), by the system (100), the predictions to generate metamodel training datasets and metamodel testing datasets;
training (314), by the system (100), the metamodel using the metamodel training datasets, and evaluating the metamodel using the metamodel testing datasets; and
assessing (316), by the system (100), performance of the metamodel, by determining one or more evaluation metrics of the metamodel, to enhance accuracy and generalization of the metamodel.
2. The method (300) as claimed in claim 1, wherein the method (300) comprises selecting (318), by the system (100), the plurality of base classifiers and the plurality of base regressors from the classifier set and the regressor set dynamically based on one or more characteristics of the dataset.
3. The method (300) as claimed in claim 1, wherein upon collecting the predictions from each of the plurality of base classifiers and each of the plurality of base regressors, the method (300) comprises storing (320), by the system (100), the predictions as train data predictions and test data predictions in a database.
4. The method (300) as claimed in claim 1, wherein for shuffling, by the system (100), the predictions, the method (300) comprises selectively transferring (322), by the system (100), a randomly selected subset of testing data into the training dataset, or a predefined set of training data into the testing dataset.
5. The method (300) as claimed in claim 1, wherein the method (300) comprises enabling (324), by the system, the metamodel to capture underlying patterns of the dataset by training and evaluating the metamodel using the metamodel training datasets and the metamodel testing datasets, respectively.
6. The method (300) as claimed in claim 1, wherein the method (300) comprises assessing (326), by the system, the performance of the metamodel using a Mean Squared Error (MSE) metric.
7. The method (300) as claimed in claim 1, wherein the one or more evaluation metrics comprise at least one of: a precision, a recall, and a F1-score.
8. A system (100) for enhancing performance of a metamodel, the system (100) comprising:
a processor (170); and
a memory (130) operatively coupled with the processor (170), wherein the memory (130) comprises one or more instructions which, when executed, cause the processor (170) to:
partition a dataset into a training dataset (trB) and a testing dataset (tsB);
upon partitioning the dataset, select at least one of: a plurality of base classifiers and a plurality of base regressors from a classifier set and a regressor set, respectively;
train each of the plurality of base classifiers or each of the plurality of base regressors on the training dataset;
evaluate each of the plurality of base classifiers or each of the plurality of base regressors on the testing dataset;
collect predictions from each of the plurality of base classifiers or each of the plurality of base regressors trained and evaluated on both the training dataset and the testing dataset;
shuffle the predictions to generate metamodel training datasets and metamodel testing datasets;
train the metamodel using the metamodel training datasets, and evaluate the metamodel using the metamodel testing datasets; and
assess performance of the metamodel, by determining one or more evaluation metrics of the metamodel, to enhance accuracy and generalization of the metamodel.
9. The system (100) as claimed in claim 8, wherein the processor (170) is configured to select the plurality of base classifiers and the plurality of base regressors from the classifier set and the regressor set dynamically based on one or more characteristics of the dataset.
10. The system as claimed in claim 8, wherein the processor (170) is configured to shuffle the predictions by selectively transferring a randomly selected subset of testing data into the training dataset, or a predefined set of training data into the testing dataset.