Loading...
Publication
OCPM4Orders – Plataforma digital de análise e rastreamento de processos de encomendas, com técnicas de Process Mining
| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 4.51 MB | Adobe PDF |
Authors
Abstract(s)
Este trabalho de projeto descreve o desenvolvimento de uma solução tecnológica baseada
em Object-Centric Process Mining (OCPM), aplicada ao contexto operacional
da Prozis, multinacional de referência na nutrição desportiva, comércio eletrónico e
logística. O trabalho cobre todo o ciclo de vida de um sistema de informação, desde
a análise de requisitos e problemas existentes, até à conceção, implementação, testes
e validação em ambiente real de produção.
A motivação principal resulta da elevada complexidade dos processos logísticos da
Prozis, onde coexistem milhões de encomendas, produtos e embalagens interligados,
processados numa infraestrutura altamente automatizada e orientada a dados. Os
sistemas de monitorização tradicionais, centrados em casos únicos, revelaram-se
insuficientes para lidar com a natureza multi-entidade das operações, dificultando a
deteção de anomalias como falhas de entrega, produtos danificados ou problemas
de armazenamento identificados via sensores Internet of Things (IoT) (temperatura,
humidade e impacto). Neste cenário, tornou-se evidente a necessidade de uma
abordagem mais expressiva que captasse as interações entre múltiplos objetos em
simultâneo.
A solução desenvolvida assenta numa arquitetura modular de três camadas
(apresentação, lógica de negócio e persistência), suportada por um modelo de dados
em estrela e alimentada por um pipeline Extract, Transform, Load (ETL) escalável
capaz de converter dados operacionais em logs compatíveis com o padrão Object-
Centric Event Log (OCEL). Entre as contribuições, destaca-se a criação de um
módulo de análise de variantes centrado em objetos, com visualizações interativas,
filtros multi-entidade e metáforas gráficas inspiradas em “mapas de metro”.
A implementação seguiu metodologias ágeis (Scrum), complementadas pelo modelo
Design Science Research, garantindo alinhamento com os objetivos empresariais
e rigor científico. Os testes de desempenho com dados reais demonstraram ganhos
significativos: redução superior a 80% no tempo médio de resposta e poupança de
cerca de 40% em memória, face a soluções tradicionais. O sistema, validado em
contexto empresarial, provando ser escalável, reutilizável e aplicável a cenários reais
de análise de processos multi-objeto.
This project work presents the development of a technological solution based on OCPM, applied to the operational context of Prozis, a leading multinational in sports nutrition, e-commerce, and logistics. The work spans the entire lifecycle of an information system, from requirements analysis and problem identification to design, implementation, testing, and validation in a production-like environment. The motivation arises from the inherent complexity of Prozis’s logistics chain, which handles over ten million yearly orders, involving interconnected products and packaging, managed through a highly automated and data-driven infrastructure. Existing monitoring systems, largely based on dashboards and spreadsheets, fail to capture the multi-entity nature of real operations, making it difficult to track anomalies such as damaged goods, failed deliveries, or sensor-based alerts (temperature, humidity, and impact) in real time. Therefore, a more expressive analytical approach was required to represent and explore the interactions of multiple entities simultaneously. The proposed solution relies on a modular three-layer architecture (presentation, business logic, and persistence), supported by a star-schema data model and powered by a scalable ETL pipeline capable of generating event logs compliant with the OCEL standard. One of its main contributions is the development of an object-centric variant analysis module, featuring interactive visualizations, multi-entity filters, and innovative visual metaphors such as subway-map representations. The implementation followed agile methodologies (Scrum), complemented by the Design Science Research framework to ensure both practical relevance and scientific rigor. Performance tests conducted with real logistics data demonstrated substantial improvements: over 80% reduction in average response time and around 40% memory savings compared to traditional approaches. Validated in a real operational environment, the system has proven to be scalable, reusable, and applicable to real-world multi-object process analysis scenarios.
This project work presents the development of a technological solution based on OCPM, applied to the operational context of Prozis, a leading multinational in sports nutrition, e-commerce, and logistics. The work spans the entire lifecycle of an information system, from requirements analysis and problem identification to design, implementation, testing, and validation in a production-like environment. The motivation arises from the inherent complexity of Prozis’s logistics chain, which handles over ten million yearly orders, involving interconnected products and packaging, managed through a highly automated and data-driven infrastructure. Existing monitoring systems, largely based on dashboards and spreadsheets, fail to capture the multi-entity nature of real operations, making it difficult to track anomalies such as damaged goods, failed deliveries, or sensor-based alerts (temperature, humidity, and impact) in real time. Therefore, a more expressive analytical approach was required to represent and explore the interactions of multiple entities simultaneously. The proposed solution relies on a modular three-layer architecture (presentation, business logic, and persistence), supported by a star-schema data model and powered by a scalable ETL pipeline capable of generating event logs compliant with the OCEL standard. One of its main contributions is the development of an object-centric variant analysis module, featuring interactive visualizations, multi-entity filters, and innovative visual metaphors such as subway-map representations. The implementation followed agile methodologies (Scrum), complemented by the Design Science Research framework to ensure both practical relevance and scientific rigor. Performance tests conducted with real logistics data demonstrated substantial improvements: over 80% reduction in average response time and around 40% memory savings compared to traditional approaches. Validated in a real operational environment, the system has proven to be scalable, reusable, and applicable to real-world multi-object process analysis scenarios.
Description
Keywords
Object-Centric Process Mining (OCPM) Directly-Follows Graphs (DFG) Process Variant Exploration Process Cubes Interactive Process Visualization Intelligent Logistics / Smart Package Platform