All Projects → matthagy → titanic-families

matthagy / titanic-families

Licence: other
Determine the families in the Kaggle Titanic competion

Programming Languages

OpenEdge ABL
179 projects
python
139335 projects - #7 most used programming language

Determine the families in the Kaggle Titanic competition

Heuristic-based algorithm to find the family trees in the titanic data.

Only involves a few parameters, all of which are related to age (e.g. minimum age for marriage). See the constant defined at the top for all of them. Neither iterative nor stochastic methods are used.

First a graph of individuals is constructed where the edges represent shared last names. This includes any previous names such maiden names. Each edge represents as relationship that can be classified as one of the following:

  • Spouse
  • Parent/Child
  • Sibling
  • Extended (e.g. aunt, cousin, or distant relative)

The classification scheme is optimistic, i.e. we only ask whether or not the relationship is possible. Much of information can be directly inferred from the given attributes (e.g. two individual cannot be siblings if one of them has sibsp==0).

Next we prove spousal relationship. This fairly easy, epically as many spouses name pairs are of the form:

West, Mrs. Edwy Arthur (Ada Mary Worth)
West, Mr. Edwy Arthur

We don't require names of this style and can also use age differences, requires Mrs title for the female, and other simple heuristics. The only difficulty arises when one individual could be classified as married to multiple individuals. There are only a few such situations and they can all be handled by assigning marriage to the couple in which the female has the males first name (e.g. Mrs. Edwy Arthur).

With spousal relationships found it is then straightforward to workout parent/child relationships. The only ambiguities at this point are child vs. sibling and they can be resolved by checking for common parent(s). Lastly, parent/child relationships can be used to work out sibling relationships. We can then recover the structure of nuclear families: families in which there is at least one parent and one or more children.

Outside of the nuclear family structure, we still maintain the relationship graph which allows for such classifications as:

  • siblings traveling together without any parents
  • extended relations
  • families joined by extended relationships

At the moment there are still some edge cases. In particular, the largest relationship graph component isn't separated into a family structure. Additionally, it would be nice to remove or relax the few parameters.

Project is licensed under the BSD (2-clause) license (see include LICENSE file)

Note that the project description data, including the texts, logos, images, and/or trademarks, for each open source project belongs to its rightful owner. If you wish to add or remove any projects, please contact us at [email protected].