Model for Managing Large Volumes of Data (MMLVD): An Approach Based on MapReduce and Machine Learning

The exponential growth of data poses challenges derived from the
5Vs of Big Data (Volume, Velocity, Variety, Veracity, and Value), which require
advanced storage, processing, and analysis models. This study presents a meta-
analysis of more than 100 publications indexed in Scopus, IEEE, and Springer
on the use of MapReduce and machine learning (ML) in Big Data management,
identifying trends, limitations, and opportunities. The results show critical short-
comings in traditional approaches: 1) high latency in iterative ML algorithms
due to repeated disk operations in MapReduce/Hadoop; 2) lack of native inter-
operability between NoSQL databases (such as MongoDB) and ML pipelines,
creating silos and costly migrations; 3) inefficient workflows that do not optimise
the entire data cycle or adequately handle speed and variety in distributed envi-
ronments. To overcome these limitations, MMLVD (Model for managing large
volumes of data) is proposed, a hybrid framework that integrates: 1) MapReduce
parallelism, 2) the predictive capabilities of machine learning, and 3) MongoDB’s
native Aggregation Framework. MMLVD optimises everything from ingestion to
predictive modelling, achieving higher performance, scalability, and cost reduc-
tion. It provides a practical and holistic solution applicable to healthcare, finance,
e-commerce, and the Internet of Things (IoT), closing the gaps left open by current
literature that relies on fragmented tools or partial evolutions such as Spark.
This model advances the state-of-the-art through a workflow-centric architecture that
natively combines distributed processing, flexible storage, and advanced predictive
analytics.