participants
Domestic Violence and Sexual Assault Prevention Center of Taichung City Government
Ever since the legislation of domestic violence prevention, every public sector has taken its own role to form a network, including Police, Social Affair, Health, Medical Care, Education, Education institutes, Judicial departments and so on.
To offer proper and complete services, Taichung City Government has integrated interdisciplinary resources from the public and private sectors to achieve zero violence and help victims elevate the ability to protect themselves. Domestic Violence and Sexual Assault Prevention Center provides diverse services including domestic violence cases handling, protection orders enactment, domestic violence offender assessment for protection orders issuing, as well as counseling service for domestic violence respondent.
When a domestic violence case was reported or a placement application was received, staff of Domestic Violence and Sexual Assault Prevention Center will perform case investigations and evaluations to make follow-up proper arrangement.
Due to case complexity and time pressure, staff with different experiences may come up with various decisions. To improve the quality and efficiency of decision making, the center provided information of long-term placement institutes, including background conditions concerning the case, situations that may encounter, regulatory requirements, and other relative records with a hope to build a model as the reference for future long-term placement institutes decisions.
The competition will start at 12:00 AM Taiwan Standard Time (UTC+8) on 2018/10/01, and end at 11:59 PM on 2018/12/31. The result will be announced on 2019/01/11.
After the participants prepare the forecast category rate of long-term placement institutes, the system would process them in batches regularly to calculate the score.
Evaluations are conducted by extracting predictive probability that is in accurate category of selected test samples, thus calculating the sum of the logarithms and the final mean value.
The formula is as follows,
$$ logloss \; = \; - \frac{1}{N} \; { \sum_{i=1}^{N} \; { \sum_{j=1}^{M} \; { y_{ij} \log({p_{ij}}) }}} $$
N: Number of samples
M: Number of categories
Yij: Whether it is an early-warning category
Pij: The predicted probability of certain category of selected samples
