Publicação
Box-Jenkin’s Methodology in Python for Stock Managing
| Resumo: | At the end of 2019, the world had shaken when social media communicated that a potential worldwide pandemic might be beginning. Early in 2020, most countries worldwide affected by the pandemic declared a state of emergency, announcing that people could not leave their houses. When confronted with these security policies, many companies faced new management challenges regarding physical and technological resources. Companies had to adapt their work style, allowing its employees to work remotely (some companies even adopted a hybrid work when the restrictions ended/ where on a break), on the other they had to adapt its technological resources for the information to be accessible for every employer with safety. For this purpose, large companies had to spend thousands or millions quickly adapting its information systems – both for acquiring more potent virtual private network and improve their capacity in terms of the online channel integration and invoice systems – as this was the only available channel to buy non-essential goods. This thesis addressed the possibility of using Machine Learning (ML) to build a predictive model to forecast which will be the sales behavior over time, by analysing a time series. That possibility consists in building a model for stock managing, that would be updated daily (with the sales till the previous day), and automatically predicts the future sales behavior, allowing an automated stock management process – not only without shortages but also without overstocking. As a result, it was achieved a fully automated ML model, using a S3 bucket (from amazon web services) connected to a Databricks instance (launched through the S3 bucket), that has the capacity to receive the sales daily, treat the data and forecast the future data points of this sales time series. |
|---|---|
| Autores principais: | Gomes, Manuel Romão dos Santos |
| Assunto: | Time Series Machine Learning Predictive Models Box-Jenkin’s Methodology Stock Managing |
| Ano: | 2022 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade Nova de Lisboa |
| Idioma: | inglês |
| Origem: | Repositório Institucional da UNL |
| Resumo: | At the end of 2019, the world had shaken when social media communicated that a potential worldwide pandemic might be beginning. Early in 2020, most countries worldwide affected by the pandemic declared a state of emergency, announcing that people could not leave their houses. When confronted with these security policies, many companies faced new management challenges regarding physical and technological resources. Companies had to adapt their work style, allowing its employees to work remotely (some companies even adopted a hybrid work when the restrictions ended/ where on a break), on the other they had to adapt its technological resources for the information to be accessible for every employer with safety. For this purpose, large companies had to spend thousands or millions quickly adapting its information systems – both for acquiring more potent virtual private network and improve their capacity in terms of the online channel integration and invoice systems – as this was the only available channel to buy non-essential goods. This thesis addressed the possibility of using Machine Learning (ML) to build a predictive model to forecast which will be the sales behavior over time, by analysing a time series. That possibility consists in building a model for stock managing, that would be updated daily (with the sales till the previous day), and automatically predicts the future sales behavior, allowing an automated stock management process – not only without shortages but also without overstocking. As a result, it was achieved a fully automated ML model, using a S3 bucket (from amazon web services) connected to a Databricks instance (launched through the S3 bucket), that has the capacity to receive the sales daily, treat the data and forecast the future data points of this sales time series. |
|---|