How to improve score


#1

Hi,
I am new to analytics.
I built a simple linear regression model and uploaded my solution. My current score as per leadership board is 1175.
I am unable to improve it.
Please suggest me , what should i do next (after linear regression) in order to improve my model and the score on leadership board :slight_smile:

Thanks.


#2

Hi @Rishabh0709, refer this blog for a deeper understanding of the solution.


#3

Hi @Rishabh0709,

Even I am very new to this and I was applying the same approach of simple linear regression taking MRP, Outlet_Type as the independent variables. I was just wondering what approach did you use to get to that score as I am getting a score of 1202 . It would be helpful if you could share some information on your approach :slight_smile:


#4

I am new to this field . tried my hand on Bigmart. When i upload my file i get an error as β€œnan”.
Can somebody plx help me?

Thanks


#5

Hi, Could you please post the first five lines of your solution file?

Also, did you refer the article mentioned above?


#6

hi! @jalFaizy

The first few lines of my solution file is

I did refer to the article above and my code is in R

Thanks


#7

Hi Rishabh,

Can you share your model that got you 1175, i am stuck with 1200s

Thanks,
Amit


#8

Hi All,
I got struck with 1148.19 using linear regression, I am trying xgboost now but now able to improve the score. I have tried lots of parameter tuning but still not able to achieve better score. Can someone please suggest how to use xgboost properly?


#9

@Shubham219 You can refer to this thread My Solution - score 1080


#10

I have already checked this. He used ensembling of gbm and neural network. I am new to data science, heard a lot about xgboost so just trying with it. But Not able to beat linear regression. After doing little bit feature engineering and feature selection I find out the following predictors which I am using.

Item_Sales/Item_MRP(Item_Sold) ~ Outlet_Location_Type+Outlet_Size+Outlet_Type+Year


#11

@Shubham219 XGBoost can be a bit difficult to tune. You can refer this guide for an overview