• a8_908209
  • 633.9KB
  • zip
  • 0
  • VIP专享
  • 0
  • 2022-06-12 02:19
bk_menu 聚类汉堡王菜单 聚类是一种无监督的学习技术,可以找到数据的结构。 非正式地,这是一种在感兴趣的对象之间找到自然分组的方法。 在这里,我使用各种聚类技术根据其营养价值对Burger King菜单项进行聚类,例如K-均值和层次聚类,高斯混合和Dirichlet过程混合基于模型的聚类,光谱聚类和基于图的基于Infomap的社区检测。 我已在其网站上以PDF格式下载了汉堡王完整菜单的最新版本,您可以在找到并将其转换为Excel。 如您所见,菜单上的每个项目都已经分组在其预定义的类别下。 例如,所有Whopper三明治都被归类为“ Whopper三明治”,而所有与儿童餐相关的菜单项(例如苹果片和儿童燕麦片)都属于“儿童餐”类别。 但是这种分组实际上并没有帮助,特别是如果您是一个健康意识强的人,想要选择一个更加健康的菜单项时。 将苹果片和水果和燕麦片归为早餐也更有意义,而不是将它
  • Clustering-Burger-King-Menu-master
  • bk_menu.Rproj
  • src
  • bk_clustering.R
  • img
  • bk_nutrition.pdf
  • sample_ggplot.png
  • data
  • bk_nutrition.csv
  • .gitignore
# bk_menu Clustering Burger King Menu Clustering is an unsupervised learning technique to find structure in data; informally, it is a way to find natural groupings among objects of interest. Here I use various clustering techniques to cluster Burger King menu items based on their nutritional values such as K-Means and Hierarchical Clustering, Gaussian Mixture and Dirichlet Process Mixture Model-Based Clustering, Spectral Clustering and Graph-Based Community Detection with Infomap. I downloaded the recent version of full Burger King menu in PDF at its website which you can find [here](img/bk_nutrition.pdf) and converted it to an Excel [file](data/bk_nutrition.csv). As you can see, each and every item on the menu is already grouped under its predefined category. For example, all the Whopper sandwiches are classified as "Whopper Sandwiches" and all the kids meal related menu items such as apple slices and kids oatmeal are under the "Kids Meals" category. But this kind of grouping is really not helpful, especially if you are one of the health-conscious who want to choose a menu item that is considerably healthier. It also makes more sense to group apple slices under fruits and oatmeal under breakfast instead of just lumping them all in kids meal section. Clustering Burger King menu items by nutritional values may reveal a different kind of groups with perhaps new surprising insights that may be more valuable for the ever growing health-conscious public. I want to know if there are menu items to certainly avoid without having to manually go through every item and analyze its nutritinoal value. Ideally I want the resulting clusters to be clear and robust, keeping in mind the number of clusters can grow with the size of the data, the number of menu items. Please refer to my [blog]( for more information about the process and results of clustering. In the blog, I came up with 7 robust clusters using Dirichlet Process Mixture Model-Based Clustering in Python that were identified as follows: * **Cluster 1**: Breakfast * **Cluster 2**: Chicken Burgers and Crispy Food * **Cluster 3**: Large Size Burgers * **Cluster 4**: Sauces, Coffees and Less Sugary Drinks * **Cluster 5**: The Ultimate Breakfast Platter (1 Item Cluster) * **Cluster 6**: Desserts and Sugary Drinks * **Cluster 7**: Milkshakes For the other clustering techniques, written here in R, I came up with around 13 clusters. K-Means and Hierarchical Clustering are fast and simple to implement but the clusters are not as robust and rather static. Gaussian Mixture Model-Based Clustering does not really give coherent clusters because Gaussian assumption does not fit well to the data. Infomap returns one huge cluster of all menu items because the adjacency matrix created from the scaled data is a network of menu items that are all too closely related which is not helpful. It's also an algorithm that attempts to find communities within a network of graphs, a sort of different problem than clustering. After cluster generation, I create a bar graph of select representative menu items from each cluster such as the Breakfast cluster here: ![ScreenShot](/img/sample_ggplot.png)
    • 聚类
    • 烧瓶聚类
    • 聚类
      Chipotle聚类挑战 语境 项目名称:Chipotle聚类挑战项目背景:BeCode,列日校区,人工智能/数据运营商训练营,2021年2月该项目的目标: 能够使用geopandas,matplotlib(和seaborn)在地图上可视化群集数据。 能够...
    • K均值聚类
    • 聚类
    • 聚类
    • 邻里聚类
    • 考试聚类
    • 聚类
    • 聚类测试