Mental health in the tech industry
-A look at data from 2014 of a ongoing survey giving a inside look at how mental health is treated in the tech industry-
It’s the year 2020 and it’s safe to say things have been going bad. The entire world is in the midst of dealing with a pandemic more severe than anything we could have perceived. A fast spreading lethal virus has spread across the globe, infecting every nation with a plague severe enough to shut down entire countries and ruin lifes across the world. The effects of this virus will undoubtedly be present in our minds and culture until the end of time.
*About 25% of American adults suffer from a diagnosable mental disorder in any given year. Mental health problems affect 1 in 5 young people, but two-thirds of those who need treatment do not have access to it.
The survey I found had just over 1,200 participants, mostly men in the Tech industry, for who had taken the survey a vast majority say there mental health has effected their work in some way. In this post I will analyse everything in the survey, to make predictions if a person who took the survey has a problem that is interfering with there work based on how they answered the other questions. Mental health problems are difficult to define not always is something you would consider a issue something that should be seen as sever. Even those diagnosed with a chronic problem should not be stigmatized in the workplace. Lots of us might have underling problems that can slightly effect or alter our overall well being and that is no different in a academic job field such as technology.
The data i used was just a fraction of the total questions in the larger survey but that left the challenge of implementing the perfect predictive model.
I chose a logistic regression for a liner model one theme i had found throughout my dig in python is that there is one factor with a much larger impact on the computer coming to its conclusion. As you might guess it was if you the have seeked treatment before. Yes having a large impact on determining a positive result and no for a negative. All together my regression model got me eighty one percent accuracy with my tree based model doing slightly better.
The treatment question ended up having about a 10 percent positive impact on my model leaving results without the treatment question it would be at around 70%. Still surprisingly high with a lot of impact coming from family history and age. Family history is a good indicator of a positive result and age for a negative.
My random Forrest accuracy was jumping around a lot when re-ran anywhere between 79–85% With gradient boost classifier It was locked into 82%. I was glad to see that the survey was split down the middle as in being taken by those whom have, and haven’t needed treatment before. I did not really know what i was getting the first time I saw this data set on kaggle it seemed to be quite a lengthy survey in its full form.
I find research into subjects of mental health or ones well being to be useful for sure. There are a lot of eye opening statistics and really they show how bad health is in America. 9% of Americans would say they suffer some level of depression. In reality, as a individual there is not much you can do(other than seek medical help). This survey is geared towards the work place. I would suggest that employers do little things to boost moral and overall happiness. I think company’s should push themselves to create a good environment and in some way, work for the employees, in a since of building a place were everyone is happy and accepted.
