Scholarlink

This study explores alcohol sales forecasting in the Ecuadorian cities of Guayaquil and Quito by combining historical sales data with socioeconomic, calendar, and weather variables. The dataset is from 2013 to 2016 and includes information about holidays, local and national events, and the Basic Vital Basket, which reflects the economic conditions affecting consumer behavior. A new classification system was introduced to categorize events based on their geographic relevance, same city, outside city, or no event. Descriptive analysis was conducted to evaluate consumption patterns across different types of events. Using a Long Short-Term Memory (LSTM) neural network, we implemented time series forecasting models with a 60-day input window. Separated models were developed for each city to account for regional differences. The results show that alcohol sales are influenced by event type, location, and economic factors. However, the models also reveal limitations due to the historical nature of the data, suggesting that integrating more recent and diverse features would improve prediction accuracy.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Predictive Analytics to Determine Alcohol Consumption at Events between Two Cities by Implementing a Data Pipeline Feeding an LSTM Model

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Nicole Uruchima

摘要

This study explores alcohol sales forecasting in the Ecuadorian cities of Guayaquil and Quito by combining historical sales data with socioeconomic, calendar, and weather variables. The dataset is from 2013 to 2016 and includes information about holidays, local and national events, and the Basic Vital Basket, which reflects the economic conditions affecting consumer behavior. A new classification system was introduced to categorize events based on their geographic relevance, same city, outside city, or no event. Descriptive analysis was conducted to evaluate consumption patterns across different types of events. Using a Long Short-Term Memory (LSTM) neural network, we implemented time series forecasting models with a 60-day input window. Separated models were developed for each city to account for regional differences. The results show that alcohol sales are influenced by event type, location, and economic factors. However, the models also reveal limitations due to the historical nature of the data, suggesting that integrating more recent and diverse features would improve prediction accuracy.