Publicação
Forecasting and Microservice Integration for Cold Chain Temperature Management
| Resumo: | Managing the cold chain is essential for preserving perishable goods, requiring strict temperature control during storage and transportation. This work explores the integration of forecasting models for more efficient temperature management in a business context, within an IoT company specialized in supply chain monitoring. The goal is to enhance predictive capabilities to anticipate temperature fluctuations and support proactive operational decisions. Three families of models were compared: a naive baseline model based on historical averages; a seasonal ARIMA model selected through AIC minimization; and an LSTM model optimized using random search and one-factor-at-a-time tuning. Two historical temperature datasets were used: one from a Cold Room, with 5,240 values sampled every 30 minutes, and another from a Standard Freezer, with 14,307 values sampled every 5 minutes. The comparative analysis showed that the LSTM model achieved the best performance for the Cold Room dataset, which exhibited more regular cycles and lower variability (MAE ≈ 0.40 ◦C; RMSE ≈ 0.52 ◦C). In contrast, for the Standard Freezer, characterized by higher-frequency fluctuations and greater noise, the (S)ARIMA model performed better (MAE ≈ 0.86 ◦C; RMSE ≈ 1.05 ◦C), while the LSTM underperformed in this noisier environment, indicating that model performance is strongly influenced by the data-generating process, particularly the stability and variability of the time series. These results highlight that the choice of model should be aligned with the stability and variability of the time series, providing practical guidance for industrial applications. The selected models were integrated into a microservice-based architecture through the development of a new predictive microservice. This service provides forecasts for the next 48 time steps for each asset, enabling access to predictive metrics and interactive dashboards. In the future, these forecasts may also be used to implement anomaly detection systems capable of identifying significant deviations between predicted and observed values, allowing early detection of potential failures. |
|---|---|
| Autores principais: | Braz,João Miguel Nogueira |
| Assunto: | Cold chain management time series forecasting, Internet of Things (IoT) (S)ARIMA LSTM |
| Ano: | 2025 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade de Lisboa |
| Idioma: | inglês |
| Origem: | Repositório da Universidade de Lisboa |
| _version_ | 1869347426506113024 |
|---|---|
| author | Braz,João Miguel Nogueira |
| author_facet | Braz,João Miguel Nogueira |
| author_role | author |
| contributor_name_str_mv | Calha,Mário João Barata Faculty of Sciences Department of Informatics Repositório Científico de Acesso Aberto da ULisboa |
| country_str | PT |
| creators_json_txt | [{\"Person.name\":\"Braz,João Miguel Nogueira\"}] |
| datacite.contributors.contributor.contributorName.fl_str_mv | Calha,Mário João Barata Faculty of Sciences Department of Informatics Repositório Científico de Acesso Aberto da ULisboa |
| datacite.creators.creator.creatorName.fl_str_mv | Braz,João Miguel Nogueira |
| datacite.date.Accepted.fl_str_mv | 2025-01-01T00:00:00Z |
| datacite.date.available.fl_str_mv | 2026-02-09T14:35:04Z |
| datacite.date.embargoed.fl_str_mv | 2026-02-09T14:35:04Z |
| datacite.rights.fl_str_mv | http://purl.org/coar/access_right/c_abf2 |
| datacite.subjects.subject.fl_str_mv | Cold chain management time series forecasting, Internet of Things (IoT) (S)ARIMA LSTM |
| datacite.titles.title.fl_str_mv | Forecasting and Microservice Integration for Cold Chain Temperature Management |
| dc.contributor.none.fl_str_mv | Calha,Mário João Barata Faculty of Sciences Department of Informatics Repositório Científico de Acesso Aberto da ULisboa |
| dc.creator.none.fl_str_mv | Braz,João Miguel Nogueira |
| dc.date.Accepted.fl_str_mv | 2025-01-01T00:00:00Z |
| dc.date.available.fl_str_mv | 2026-02-09T14:35:04Z |
| dc.date.embargoed.fl_str_mv | 2026-02-09T14:35:04Z |
| dc.format.none.fl_str_mv | application/pdf |
| dc.identifier.none.fl_str_mv | http://hdl.handle.net/10400.5/116927 |
| dc.language.none.fl_str_mv | eng |
| dc.rights.none.fl_str_mv | http://purl.org/coar/access_right/c_abf2 |
| dc.subject.none.fl_str_mv | Cold chain management time series forecasting, Internet of Things (IoT) (S)ARIMA LSTM |
| dc.title.fl_str_mv | Forecasting and Microservice Integration for Cold Chain Temperature Management |
| dc.type.none.fl_str_mv | http://purl.org/coar/resource_type/c_bdcc |
| description | Managing the cold chain is essential for preserving perishable goods, requiring strict temperature control during storage and transportation. This work explores the integration of forecasting models for more efficient temperature management in a business context, within an IoT company specialized in supply chain monitoring. The goal is to enhance predictive capabilities to anticipate temperature fluctuations and support proactive operational decisions. Three families of models were compared: a naive baseline model based on historical averages; a seasonal ARIMA model selected through AIC minimization; and an LSTM model optimized using random search and one-factor-at-a-time tuning. Two historical temperature datasets were used: one from a Cold Room, with 5,240 values sampled every 30 minutes, and another from a Standard Freezer, with 14,307 values sampled every 5 minutes. The comparative analysis showed that the LSTM model achieved the best performance for the Cold Room dataset, which exhibited more regular cycles and lower variability (MAE ≈ 0.40 ◦C; RMSE ≈ 0.52 ◦C). In contrast, for the Standard Freezer, characterized by higher-frequency fluctuations and greater noise, the (S)ARIMA model performed better (MAE ≈ 0.86 ◦C; RMSE ≈ 1.05 ◦C), while the LSTM underperformed in this noisier environment, indicating that model performance is strongly influenced by the data-generating process, particularly the stability and variability of the time series. These results highlight that the choice of model should be aligned with the stability and variability of the time series, providing practical guidance for industrial applications. The selected models were integrated into a microservice-based architecture through the development of a new predictive microservice. This service provides forecasts for the next 48 time steps for each asset, enabling access to predictive metrics and interactive dashboards. In the future, these forecasts may also be used to implement anomaly detection systems capable of identifying significant deviations between predicted and observed values, allowing early detection of potential failures. |
| dirty | 0 |
| eu_rights_str_mv | openAccess |
| format | masterThesis |
| fulltext.url.fl_str_mv | https://repositorio.ulisboa.pt/bitstreams/c0f09dd7-5b51-4f96-9314-84b5dde490d9/download |
| id | ul_518c00a861403612aabb46a424dfafbf |
| identifier.url.fl_str_mv | http://hdl.handle.net/10400.5/116927 |
| inst_facet_str | urn:organizationAcronym:ul{{{_:::_}}}Universidade de Lisboa |
| instacron_str | ul |
| institution | Universidade de Lisboa |
| instname_str | Universidade de Lisboa |
| language | eng |
| network_acronym_str | ul |
| network_name_str | Repositório da Universidade de Lisboa |
| oai_identifier_str | oai:repositorio.ulisboa.pt:10400.5/116927 |
| organization_str_mv | urn:organizationAcronym:ul |
| person_str_mv | Braz,João Miguel Nogueira |
| publishDate | 2025 |
| repo_facet_str | urn:repositoryAcronym:ul{{{_:::_}}}Repositório da Universidade de Lisboa |
| reponame_str | Repositório da Universidade de Lisboa |
| repository_id_str | urn:repositoryAcronym:ul |
| service_str_mv | urn:repositoryAcronym:ul |
| spelling | engenManaging the cold chain is essential for preserving perishable goods, requiring strict temperature control during storage and transportation. This work explores the integration of forecasting models for more efficient temperature management in a business context, within an IoT company specialized in supply chain monitoring. The goal is to enhance predictive capabilities to anticipate temperature fluctuations and support proactive operational decisions. Three families of models were compared: a naive baseline model based on historical averages; a seasonal ARIMA model selected through AIC minimization; and an LSTM model optimized using random search and one-factor-at-a-time tuning. Two historical temperature datasets were used: one from a Cold Room, with 5,240 values sampled every 30 minutes, and another from a Standard Freezer, with 14,307 values sampled every 5 minutes. The comparative analysis showed that the LSTM model achieved the best performance for the Cold Room dataset, which exhibited more regular cycles and lower variability (MAE ≈ 0.40 ◦C; RMSE ≈ 0.52 ◦C). In contrast, for the Standard Freezer, characterized by higher-frequency fluctuations and greater noise, the (S)ARIMA model performed better (MAE ≈ 0.86 ◦C; RMSE ≈ 1.05 ◦C), while the LSTM underperformed in this noisier environment, indicating that model performance is strongly influenced by the data-generating process, particularly the stability and variability of the time series. These results highlight that the choice of model should be aligned with the stability and variability of the time series, providing practical guidance for industrial applications. The selected models were integrated into a microservice-based architecture through the development of a new predictive microservice. This service provides forecasts for the next 48 time steps for each asset, enabling access to predictive metrics and interactive dashboards. In the future, these forecasts may also be used to implement anomaly detection systems capable of identifying significant deviations between predicted and observed values, allowing early detection of potential failures.application/pdfenForecasting and Microservice Integration for Cold Chain Temperature ManagementBraz,João Miguel NogueiraCalha,Mário João BarataFaculty of SciencesDepartment of InformaticsHostingInstitutionOrganizationalRepositório Científico de Acesso Aberto da ULisboae-mailmailto:repositorio@ulisboa.ptrepositorio@ulisboa.ptURNurn:tid:2041737792026-02-09T14:35:04Z20252025-01-01T00:00:00ZHandlehttp://hdl.handle.net/10400.5/116927http://purl.org/coar/access_right/c_abf2open accessCold chain managementtime series forecasting,Internet of Things (IoT)(S)ARIMALSTM4117770 bytesliteraturehttp://purl.org/coar/resource_type/c_bdccmaster thesishttp://purl.org/coar/access_right/c_abf2application/pdffulltexthttps://repositorio.ulisboa.pt/bitstreams/c0f09dd7-5b51-4f96-9314-84b5dde490d9/download |
| spellingShingle | Forecasting and Microservice Integration for Cold Chain Temperature Management Braz,João Miguel Nogueira Cold chain management time series forecasting, Internet of Things (IoT) (S)ARIMA LSTM |
| status | SINGLETON |
| subject.fl_str_mv | Cold chain management time series forecasting, Internet of Things (IoT) (S)ARIMA LSTM |
| title | Forecasting and Microservice Integration for Cold Chain Temperature Management |
| title_full | Forecasting and Microservice Integration for Cold Chain Temperature Management |
| title_fullStr | Forecasting and Microservice Integration for Cold Chain Temperature Management |
| title_full_unstemmed | Forecasting and Microservice Integration for Cold Chain Temperature Management |
| title_short | Forecasting and Microservice Integration for Cold Chain Temperature Management |
| title_sort | Forecasting and Microservice Integration for Cold Chain Temperature Management |
| topic | Cold chain management time series forecasting, Internet of Things (IoT) (S)ARIMA LSTM |
| topic_facet | Cold chain management time series forecasting, Internet of Things (IoT) (S)ARIMA LSTM |
| url | http://hdl.handle.net/10400.5/116927 |
| visible | 1 |