Applications
See how Kumbi works.
You can find out which McDonald's menu is the most suitable for you, save on GSM subscription or explore the dependencies between the statistics describing Poland. And much more.
McDonald menu
The McDonalds menu application allows you to choose from the McDonalds restaurant menu products with your preferred content of calorie, protein, carbohydrate and salt. In addition, you can choose products that do not contain harmful allergens.
The main difference compared to traditional comparisons technology is that the Kumbi algorithm compares the content of all components simultaneously. The algorithm selects products containing the most similar amounts and the proportions of all components (assessment criteria).
The data comes from the McDonald website.The data does not constitute a commercial offer.
Data from June 2017.
GSM subscription
With this application you can select a GSM subscription according to number of criteria at the same time. You can select the best for you offer based on eg price, period of notice, free social media, Internet transfer and many more.
The GSM Subscription model contains data about most GSM suppliers available on the market.
The data comes from the suppliers websites, so they may be out of date and/or be inaccurate.
The data consists of basic offers. Offers do not include discounts, promotions and offers with telephones.
Data are presented for information purposes only and are not a commercial offer.
Data from June 2017.
GSM Starters
With this application you can select a GSM starter according to number of criteria at the same time. You can select the one best for you based on eg price, period of notice, free social media, Internet transfer and many more.
The GSM Starter model contains data about most GSM suppliers available on the market.
The data comes from the suppliers websites, so they may be out of date and/or be inaccurate.
The data consists of basic offers. Offers do not include discounts, promotions and offers with telephones.
Data are presented for information purposes only and are not a commercial offer.
Data from June 2017.
Satelite TV
The Satelite TV model contains offers of NC+ and Polsat.
The data comes from the suppliers websites, so they may be out of date and/or be inaccurate.
The data consists of basic offers. Offers do not include discounts and promotions.
Data are presented for information purposes only and are not a commercial offer.
Data from September 2017.
Maps
The application sorts statistics. It determines degree of similarity of a statistics to the selected one. Similarity is calculated based on statistics distribution into administrative regions.
This example shows how the algorithm compares objects. In this case statistics. Notice that it compares not only values but mainly structure of the criteria. Degree of similarity is high when statistics have similar distribution into regions but different values in corresponding regions. But similarity degree is equal 100% (are identical) only when al values are the same, so the structure of criteria is the same by definition.
This model calculates degree of relative similarity of the statistics distribution into provinces.
Relative similarity is the degree of similarity between statistics contained in the database. This means that the degree of similarity of the most similar statistics in the database (in practice the one to which the others are compared) is 100% and for the one that is the least similar, the degree of similarity is 0%.
In this version of the application, absolute similarity is not calculated. Absolute similarity is calculated based on theoretically the least similar solution (object) for which the degree is equal 0%.
Not in this simple model byt there is possible to give also weights to provinces to modify their influence on the degree of similarity.
The database contains currently 240 statistics published by the Central Statistical Office (GUS), Local Data Bank and the National Electoral Commission (PKW).
Training Model
Here you can play with Kumbi algorithm to get familiar with it
The training model contains data of cuboids. Data consists of height, length, width, weight and price. You can choose a cuboid with the most desirable dimensions, weight and price.
Start with a single criterion, e.g. length.
At the beginning, as Expected Value use only MAX or MIN and as Weights valuees 100% to all criteria. This way it is much easier to understand how Kumbi calculates the similarity.
Introduce further criteria gradually. Change Expected Values and Weights. Check if the results match your expectations.
You can also test how your intuition works. Try to predict the results. Write them down. And next try to choose Weights so that you get the expected result. A great exercise to understand how Kumbi works and ... your intuition.
Kumbi
