System and method for generating a predictive model based on automated data modelling

1 . A method for generating a predictive model, wherein the method comprising:

receiving, by a processor, data, wherein the data comprises a plurality of features, wherein the plurality of features is indicative of columns in the data;

extracting, by the processor, one or more predicator features from the data based on a feature selection methodology, wherein the feature selection methodology comprises:

computing, by the processor, a degree connectedness for each of the plurality of features using a modified mutual information technique and a Pearson correlation co-efficient value,

wherein the degree connectedness is indicative of a connection between two or more features from the plurality of features,

wherein the modified mutual information technique computes a non-linear relationship between at least two features from the plurality of features for the computation of the degree connectedness, and

wherein the Pearson correlation co-efficient is used to compute a linear relationship between the at least two features from the plurality of features for the degree connectedness;

identifying, by the processor, the one or more predicator features on a comparison of degree of connectedness and a predefined threshold; and

executing, by the processor, a data cleansing, wherein the data cleansing comprises:

deleting, by the processor, constant data from the plurality of features, wherein the constant data from the plurality of features do not provide information related to the one or more predicator features;

computing, by the processor, a Pearson correlation coefficient value between each of the plurality of feature;

eliminating, by the processor, one or more features from the plurality of features based on similar Pearson correlation coefficient value;

detecting, by the processor, one of an anomaly and an outliers in the data based on a statistical distribution analysis methodology; and

replacing, by the processor, one of the anomaly and the outliers by one of the descriptive statistics;

identifying, by the processor, a data type associated with the data and the one or more predicator features, wherein the data type comprises a time-series data type, a non-time series data type, a numerical data type, a categorical data type, seasonal stationary data type, a stationary data type or trend stationary, wherein the stationary or the trend stationary data type is identified using a Augmented Dickey-Fuller test, wherein the time series or the non-time series data type is identified using alternate hypothesis method, and wherein the seasonal stationary data type is identified using a Hurst Exponent Measure;

computing, by the processor, a feature data distribution for the categorical features and the non-numeric features based on a frequency encoding methodology;

generating, by the processor, a predictive model applied on the data based on the data type and the one or more predicator features, wherein predictive model comprises a classification model, a time series model and a regression model type;

modifying tuning parameters of the predictive model;

optimizing the modified predictive model using a genetic algorithm; and

applying a mappers and reducers model in combination with a message passing interface (MPI) on the predictive model and the data for iterative computations,

wherein the mappers and reducers model is configured to execute feature analysis and model generation in a distributed framework, and

wherein the MPI is configured to coordinate iterative execution and data exchange across distributed nodes within the distributed framework, thereby achieving speed and scalability for a scalable distributed implementation on the distributed framework.

2 . The method of claim 1 , wherein the feature selection methodology further comprises:

executing, by the processor, a statistical analysis, wherein the statistical analysis comprises:

classifying, by the processor, the plurality of features into one of a numeric features, a categorical features and a non-numeric features;

computing, by the processor, descriptive statistics for the numeric features, wherein the descriptive statistics comprises one or more of a minimum value, a maximum value, a mean value, a standard deviation and a standard distribution.

3 . The method of claim 1 , wherein generating a predictive model is based on federated learning methodology on the distributed framework.

4 . A system for generating a predictive model, wherein the system comprising:

a memory; and

a processor coupled to the memory, wherein the processor is configured to execute program instructions stored in the memory for:

receiving data, wherein the data comprises a plurality of features, wherein the plurality of features is indicative of columns in the data;

extracting one or more predicator features from the data based on a feature selection methodology, wherein the feature selection methodology comprises:

computing a degree connectedness for each of the plurality of features using a modified mutual information technique and a Pearson correlation co-efficient value, wherein the degree connectedness is indicative of a connection between or more features from the plurality of features,

wherein the modified mutual information technique computes a non-linear relationship between at least two features from the plurality of features for the computation of the degree connectedness, and

wherein the Pearson correlation co-efficient is used to compute a linear relationship between the at least two features from the plurality of features for the degree connectedness;

identifying the one or more predicator features on a comparison of degree of connectedness and a predefined threshold; and

executing a data cleansing, wherein the data cleansing comprises:

deleting constant data from the plurality of features, wherein the constant data from the plurality of features do not provide information related to the one or more predicator features;

computing a Pearson correlation coefficient value between each of the plurality of feature;

eliminating one or more features from the plurality of features based on similar Pearson correlation coefficient value;

detecting one of an anomaly and an outliers in the data based on a statistical distribution analysis methodology; and

replacing one of the anomaly and the outliers by one of the descriptive statistics;

identifying a data type associated with the data and the one or more predicator features, wherein the data type comprises a time-series data type, a non-time series data type, a numerical data type, a categorical data type, seasonal stationary data type, a stationary data type or trend stationary, wherein the stationary or the trend stationary data type is identified using a Augmented Dickey-Fuller test, wherein the time series or the non-time series data type is identified using alternate hypothesis method, and wherein the seasonal stationary data type is identified using a Hurst Exponent Measure;

computing a feature data distribution for the categorical features and the non-numeric features based on a frequency encoding methodology;

generating a predictive model to be applied on the data based on the data type of the one or more predicator features and the one or more predicator features, wherein predictive model comprises a classification model, a time series model and a regression model type;

modifying tuning parameters of the predictive model;

optimizing the modified predictive model using a genetic algorithm; and

applying a mappers and reducers model in combination with a message passing interface (MPI) on the predictive model and the data for iterative computations,

wherein the mappers and reducers model is configured to execute feature analysis and model generation in a distributed framework, and

wherein the MPI is configured to coordinate iterative execution and data exchange across distributed nodes within the distributed framework, thereby achieving speed and scalability for a scalable distributed implementation on the distributed framework.

5 . The system of claim 4 , wherein the feature selection methodology comprises:

executing a statistical analysis, wherein the statistical analysis comprises:

classifying the plurality of features into one of a numeric feature, a categorical features and a non-numeric features;

computing descriptive statistics for the numeric features, wherein the descriptive statistics comprises one or more of a minimum value, a maximum value, a mean value, a standard deviation and a standard distribution.

6 . The system of claim 4 , wherein generating a predictive model is based on federated learning methodology on the distributed framework.