ANN for my data

LuckyLukeLuckyLuke Member Posts: 3 Learner I
Hi all,
i need help to analyze my data set for my thesis but could not proceed per my limited knowledge about rapidminer. I have accidents taken place in construction and some attributes. I want to use root causes (15 type of root cause-15 neurons) and immediate causes (19 types of causes-19 neurons) as layers in neural networks and try to predict outcomes as event exposure types (6 different types). For example (from picture) first accident took place in 27.06.2015 @ 16:15 and resulted in Falling from height. IC3, IC10 and IC16 contributed  as immediate causes. RC4 and RC15 are root causes that led to accidents.In other words for first accident activation of neurons  (3-10-16 in first layer) activates  neurons in second layer (neurons 4 and 15) and one outcome occurs. I have many of accidents like this. See pic for illustration. I need urgent help, any help is appreciated. 

Comments

  • sgenzersgenzer Administrator, Moderator, Employee, RapidMiner Certified Analyst, Community Manager, Member, University Professor, PM Moderator Posts: 2,959 Community Manager
    hi @LuckyLuke - welcome to the community. If you're just getting started with RM, I'd recommend the RM Academy for learning the ins and outs. Neural nets are covered in the RM Academy here: https://academy.rapidminer.com/learn/video/neural-nets-demo

    Scott
  • LuckyLukeLuckyLuke Member Posts: 3 Learner I
    edited January 2020
    sgenzer 
    Thank you  for the suggestion, i went through it  but still need more help
  • lionelderkrikorlionelderkrikor Moderator, RapidMiner Certified Analyst, Member Posts: 1,195 Unicorn
    Hi @LuckyLuke,

    I'm not specialist of neural network(s) but at first sight I don't know how to handle ICF and RCF directly like that.
    From my opinion , you have first  to split :  
     - your "meta attribute" ICF into 5 attributes (called for example ICF-1, ...,ICF-5) because  according to your screenshot
    you can have up to 5 possible ICF for a given accident
     - your "meta attribute" RCF into 4 attributes (called for example RCF-1, ...,RCF-4) because  according to your screenshot
    you can have up to 4 possible RCF for a given accident.

    More over I'm not sure to understand , what is your label ? (the attribute you want to predict).

    The best is, if possible, to share your dataset. I will try to play with your data / build a model.

    Regards,

    Lionel


  • LuckyLukeLuckyLuke Member Posts: 3 Learner I
    Hi lionelderkrikor,  i want to predict event exposure type (F, T, C etc.). may i have your email to send you my data set ?
  • lionelderkrikorlionelderkrikor Moderator, RapidMiner Certified Analyst, Member Posts: 1,195 Unicorn
    Hi @LuckyLuke,

    I sent you my e-mail via private message...

    Regards,

    Lionel
Sign In or Register to comment.