Transcript:

If you can adapt to a straight line, then you can adapt to a curved curve, make me smile, make me giggle Hello, I’m Josh Stormer (Josh Stormer), you are welcome to participate in the statistical task today, we will discuss logistic regression This is a technique that can be used in traditional statistics and machine learning, so let’s get started Before delving into logistic regression, let’s get started. A, take a step back and review, linear regression We talked about another statistical task about linear regression We have some data Weight and dimensions Then we adjust it With that line, we can do many things First, we can calculate r-squared and determine whether the weight and size are A relevant large value indicates a large impact, and Second, calculate the p-value to determine whether the r-squared value is statistically significant, Third, if a new mouse has this weight, we can use lines to predict the size of a given weight This is the size we predict based on weight, even though At that time we did not use data to predict certain things belonged to the category of machine learning So plain old linear regression is a form of machine learning We also talked about multiple regression Now we are trying to use weight and blood volume to predict size Or we can say that we are trying to model size using weight and blood volume Multiple regression has the same effect as normal regression We calculated the sum of squared r We calculated the p-value, We can use weight and blood volume to predict size, This makes multiple regression a slightly better method of machine learning We also discussed how to use discrete measures (such as genotype) to predict size People who are not familiar with the term genotype will certainly not. No, it’s important, just know that it refers to different types of mice Finally, we can compare the models So, on the left, we performed a normal regression, using weights to predict the size and We can compare these predictions with the predictions we get from multiple regression, where we use weight and blood volume to predict weight Comparing simple models with complex models can tell us whether we need to measure weight and blood volume to make accurate predictions Size, or can we walk away with just weight Now, we remember all the great work that linear regression can do Let’s talk about logistic regression Logistic regression is similar to linear regression apart from Logistic regression predicts whether things are right or wrong, rather than predicting continuous things (such as sighs) These mice are obese, These mice are not Similarly, instead of fitting the logistic regression line of the data, the sigmoid logistic function is also fitted The curve changes from zero to one? Moreover, this means that the curve will tell you the likelihood of obesity based on the weight of the mouse What if we weigh a very heavy mouse? Is there a high probability that the new mice will be obese? If we weigh an intermediate mouse So only 50% of mice have a chance of being obese? Finally, light mice are unlikely to be obese Although logistic regression can tell us whether the mouse is obese, it is usually used for classification For example, if the mouse is more than 50% likely to be obese Then we classify it as obese Otherwise we will classify it as “not obese” Just like linear regression, in this case, we can build a simple model. Can we predict obesity by weight or obesity? In this case, more sophisticated models predict obesity by weight and genotype In this case, obesity is predictable. By weight, genotype, age and Finally, obesity is predicted by weight genotype, age and age In other words, just like linear regression logic, astrological symbols Regression can handle continuous data such as weight and age, as well as discrete data such as genotype and astrological signs We can also test whether each variable is useful for predicting obesity however Unlike normal regression, we cannot easily compare complex models with simple models. We will discuss in detail why Instead, we just test to see if the influence of the variable on the prediction is significantly different from zero If not, it means that the variable does not help predict We use Wald’s test to figure out what we’re going to discuss, In this case, the astrological symbol is a useless handbag That statistical term is not helpful to us This means that we can save time and space for learning. By omitting Logistic regression capability that uses continuous and discrete measurements to provide new sample probabilities and classify them Make it a popular machine learning method One big difference between linear regression and logistic regression is how the line fits the data With linear regression, we use the least squares method to fit a straight line In other words, we found the line that minimizes the sum of squares of these residuals We also use the residual to calculate r. Squaring compares simple and complex models Logistic regression does not have the same concept of residuals, so least squares cannot be used and r-squared cannot be calculated Instead, it uses what is called maximum likelihood There is an entire stack search on maximum likelihood, so check the details, but in summary You, choose the probability of scaling. By observing the weight of obese mice, like this curve and You, use it to calculate the probability of observing such a heavy non-obese mouse, Then you calculate the observed probability, this mouse and You do this to all mice Finally, multiply all these possibilities together, and this is the probability given this row of data Then you move the line and calculate the new data possibility Then transport the production line and calculate the probability again, then once again Finally, choose the curve with the greatest likelihood as bam In short, logistic regression can be used to classify samples and It can use different types of data (such as size and/or genotype) for classification and It can also be used to evaluate which variables are useful for sample classification, namely Constellations are useless tote bags Thank you, we have completed another exciting statistical task, just like StackQuest. For more information, please subscribe If you have suggestions for future statistical tasks, please put them in the comments below until next time