System and method for improving machine learning model performance in a communications network

A method operative with a mobile communications system, the method comprising:

• providing a machine learning (ML) model for a radio access network (RAN) infrastructure disposed in the mobile communications system;
• training the ML model using a modified dataset obtained for a plurality of cellular aggregation units (CAUs) forming the RAN infrastructure, wherein the modified dataset is derived from data collected for individual CAUs over a data collection period with respect to a plurality of variables and optimized by replacement of null values of a particular variable by a modal value of the particular variable;
• executing the trained ML model, responsive to a set of test data associated with the RAN infrastructure, to obtain a prediction of one or more key performance indicators (KPIs) relative to the mobile communications system;
• and
• providing one or more control inputs to at least a portion of the RAN infrastructure responsive to the one or more predicted KPIs, wherein the one or more control inputs are operative to effectuate a resource configuration adjustment with respect to the at least a portion of the RAN infrastructure.

A network element associated with a mobile communications network, the network element comprising:

• one or more processors;
• and
• one or more persistent memory modules coupled to the one or more processors, the one or more persistent memory modules having program instructions stored thereon that perform a data conditioning method operative with the mobile communications network when executed by the one or more processors of the network element, the program instructions comprising:
• a code portion for obtaining data from a plurality of cellular aggregation units forming a RAN infrastructure of the mobile communications network, the data collected for individual CAUs over a data collection period with respect to a plurality of variables;
• a code portion for separating the data into a plurality of data frames, each data frame identified by an object identifier and corresponding to a respective individual CAU;
• a code portion for performing data cleaning operations in a particular sequential order as follows:
• removing data frames that have less than a threshold number of rows for each data frame, wherein each row corresponds to a time point of the data collection period at which data for the plurality of variables is collected;
• removing sparse columns corresponding to variables from each data frame that have less than a select number of entries the data collection period;
• and
• removing data frames that have less than a select number of variables corresponding to a designated number of valid key performance indicators (KPIs), thereby obtaining a reduced dataset;
• and
• a code portion performing following operations for each remaining data frame of the reduced dataset to obtain a modified dataset therefrom:
• determining a modal value for each of remaining variables of the data frame;
• and
• if a particular variable contains a null value, replacing the null value with the modal value of the particular variable.