Inspiration
Our idea is based on the difficulties experienced by one of our friends who owns a small F&B business. From that feedback, we conducted research to find out if there was indeed a problem.
In addition to the research, we also asked our parents, who are also part of small businesses but in different industries. From the interviews we had, the results were quite similar, where they shared the same problem: they were confused about the future plans for their business.
Apart from the interviews, we also observed the fact that many SMEs in Indonesia do not utilize data effectively. Although the government now requires tax reporting, this data is often only used for formalities, especially for micro-sized SMEs. Meanwhile, there is an article from McKinsey which states that using data can increase sales by up to 60%.
Therefore, we want to create software with the help of machine learning and large language models that can assist SMEs in Indonesia in understanding their data and making decisions based on data.
What It Does
Our application has 3 main features:
Provides sales insights using machine learning.
Provides promo/bundle recommendations using an AI agent with a large language model.
Predicts demand for the upcoming week, making F&B SMEs in Indonesia more efficient with the use of machine learning.
However, we require their sales data to analyze.
What makes our application unique is:
The data we use is POS data, which is highly applicable to current F&B businesses.
We also add external data such as market trends from Google Trends and weather conditions in certain regions.
How We Built It
We use an AI agent with Dify to create workflows for promo recommendations. These recommendations are divided into 5 types: bundling, cross-selling, discount promos, and flash sales.
We use the LSTM Multivariate model, which applies time series to provide more accurate sales predictions. However, it still requires improvement because the POS data we have is only about 6 months' worth.
Initially, we will ask the user to send their sales data, and then our backend will preprocess it to be ready for input into our model.
Challenges We Ran Into
- The dataset we received is very large, about 20GB, which required us to take more time to process the data.
Accomplishments That We're Proud Of
We have successfully created a machine learning model.
We also provide data and insights that are easy to understand, so small business owners can easily comprehend them.
What We Learned
Utilize time wisely, as it is very valuable.
Clearly state our ideas.
Communication is important.
What's Next for Gastrolytics
We will continue to develop our machine learning model.
We will expand our machine learning capabilities so that it can be used for industries beyond F&B.# Inspiration
Our idea is based on the difficulties experienced by one of our friends who owns a small F&B business. From that feedback, we conducted research to find out if there was indeed a problem.
In addition to the research, we also asked our parents, who are also part of small businesses but in different industries. From the interviews we had, the results were quite similar, where they shared the same problem: they were confused about the future plans for their business.
Apart from the interviews, we also observed the fact that many SMEs in Indonesia do not utilize data effectively. Although the government now requires tax reporting, this data is often only used for formalities, especially for micro-sized SMEs. Meanwhile, there is an article from McKinsey which states that using data can increase sales by up to 60%.
Therefore, we want to create software with the help of machine learning and large language models that can assist SMEs in Indonesia in understanding their data and making decisions based on data.
What It Does
Our application has 3 main features:
Provides sales insights using machine learning.
Provides promo/bundle recommendations using an AI agent with a large language model.
Predicts demand for the upcoming week, making F&B SMEs in Indonesia more efficient with the use of machine learning.
However, we require their sales data to analyze.
What makes our application unique is:
The data we use is POS data, which is highly applicable to current F&B businesses.
We also add external data such as market trends from Google Trends and weather conditions in certain regions.
How We Built It
We use an AI agent with Dify to create workflows for promo recommendations. These recommendations are divided into 5 types: bundling, cross-selling, discount promos, and flash sales.
We use the LSTM Multivariate model, which applies time series to provide more accurate sales predictions. However, it still requires improvement because the POS data we have is only about 6 months' worth.
Initially, we will ask the user to send their sales data, and then our backend will preprocess it to be ready for input into our model.
Challenges We Ran Into
- The dataset we received is very large, about 20GB, which required us to take more time to process the data.
Accomplishments That We're Proud Of
We have successfully created a machine learning model.
We also provide data and insights that are easy to understand, so small business owners can easily comprehend them.
What We Learned
Utilize time wisely, as it is very valuable.
Clearly state our ideas.
Communication is important.
What's Next for Gastrolytics
We will continue to develop our machine learning model.
We will expand our machine learning capabilities so that it can be used for industries beyond F&B.
Built With
- chatgpt
- flask
- lstm
- machine
- machine-learning
- nextjs
Log in or sign up for Devpost to join the conversation.