Articles
-
Time-Series Forecasting using Prophet- part2
May 30, 2020 - Shoaib Khan
In the last post, we learned that we can build a simple Prophet model by combining trend and seasonality to forecast the future. We saw, that in the model equation not only can we add those factors but we can also include holidays effects and extra regressors, which makes it extremely powerful.
-
Time-Series Forecasting using Prophet- part1
May 20, 2020 - Shoaib Khan
Prophet Prophet is a time-series forecasting library by facebook and it’s available in Python and in R. It is simple to use and requires very little classical time-series experience, for e.g. ARIMA, SARIMA etc. Thus, you can start building models quickly, get good results and become effective in no time. Considering this, prophet has become a popular library for time-series applications.
-
Adaptive models during COVID-19
April 17, 2020 - Shoaib Khan
COVID-19 has derailed many aspects of life, and it’s hard to adapt. However, to survive we must. The story is hardly different when we look at our mathematical models, many good models have come crashing down as they failed to adjust to sudden fluctuations in the data.
-
Why ECDF is better than a Histogram
January 20, 2019 - Shoaib Khan
If you have never heard of binning bias then you could be using histograms and drawing incorrect conclusions. Please read my original post on Medium and it will help you avoid those mistakes.
-
Conundrums of the Confusion Matrix
January 12, 2019 - Shoaib Khan
Its extremely important to understand your model’s performance. In this post, I explain fundamental metrics such as precision, recall, accuracy, error rate etc.. these metrics help us evaluate all types of classification models e.g. logistic regression, random forest, neural network etc. I had written this post originally on Medium and its been appreciated by many for its simplicity and visuals.