Machine Learning - Data Collection and Sampling Techniques

In this video, we are going to discuss data collection and sampling techniques

So now in case of data collection, this is very important in our machine learning because

in our machine learning actually we'll be working of huge set of data

So we'll be working on huge set of data and how this data can be collected

So we shall be discussing that one and also the sampling techniques

So, data can be collected in variety of ways. So, one of the most common method is through the use of surveys

And the surveys can be done by using a variety of methods

So we can go for the telephonic surveys, we can go for the Milt Cossiniar surveys

we can go for the personal interview surveys and surveys record or direct observation of the

situation that are multiple different ways in which you can do the survey to collect our data

So, in case of telephonic survey, it is very much cheaper because you need not to move

physically and but sometimes some people may not be picking up the phone calls, they may

not be reachable, so that are the different disadvantages in telephonic survey

In case of mail the questionnaire surveys, so now it is a little bit costly because you

are supposed to mail all these questionnaires by courier or by some other means but you also can also send this particular questionnaire through the emails also But in this cases all the mails whatever the questionnaires you have sent may not be replied back

But the people might be more candid in their opinion because there is no directly you are not interrogating he or her to gather data

Personal interview surveys, in this case it is very much fruitful because you can do the survey

personally meeting with the person but some cases in some cases the

interviewer might remain biased in the co-signor set so this particular personal interview surveys may have some other flaws serving records or

direct observation of the situation so there are different ways in which the

data can be collected through this survey method so statistician use four

basic methods of sampling so here we're having the random sampling systematic

sampling stratified sampling and cluster sampling so we know that whenever our

population is a huge subjects are there in the population and this population

cannot be reached to interact with each and every subject is not feasible then

we'll be going for the sampling so sample is nothing but collection of subjects

taken from the population so at first we are going for the random samples So random samples are selected by using chance methods or might be using some random numbers So from the population we are selecting random number of subjects using the random samples

Next one is the researcher obtained systematic samples by numbering each subject of the

population and then selecting every kth subject. So let us suppose we are having, say, 100 number of, so 100 number of subjects

we are having. Out of them, we are trying to find out, we're trying to interrogate or we're

trying to pick up a sample of the size, say, 20. In that case, 100 by 20 is 5. So, we can

go for picking up the fifth subject from the population. This 120 are just used as an example

These figures are only for the example and explanation. So, researchers obtain systematic

samples by numbering each subject of the population and then selecting a numbering

every kath subject from the population. Next one we are going for the stratified

samples. So, the researcher obtained stratified samples by dividing the population into groups also known as strata according to the sum characteristic. This is

important to study and then sampling from each group. So what will happen

Depending upon some property the population will be divided into multiple number of strata and then we shall do the sampling taking taking subjects from each strata next one we are going for the

cluster samples researcher also use cluster samples here the population will be

divided into groups also known as clusters by some means such as using some

geographic location area or respective different other properties like say schools in the large school district

So, using some property, the population will be divided into multiple clusters

And then the researcher randomly select some of the clusters and use or access all the

subjects belonging to those selected clusters. So, that is, use all the members of the selected clusters at the subjects of the samples

So that is another way to do the sampling. The last one is our convenience sampling

So, here a researcher uses subjects that are convenient. For example, the researcher may interview subjects entering a local mall to determine the nature

of their visit and perhaps what stores they will be visiting for the shopping purpose

So in this way, there are different ways in which we can do the collection of data

So one method is our survey and then what are the different ways we can create one sample

from the population. Thanks for watching this video