Data cleaning and data wrangling is fundamental in Data Science. Cleaning and boosting the base allows us optimal quality results.
After this cleaning we make the exploratory analysis and find the first relationships with the information.
Infer from our data, based on a variable objective, supervised learning will help us find an answer.
Algorithms such as linear regression, logistic, Lasso, KNN, decision trees, support vector machines or neural networks will allow us to find solutions.
Find relationships, but we do not have an objective variable, through unsupervised learning we can find groups or clusters of our interest and know the behavior of our base.
Algorithms such as hierarchical clustering, k-means, k-meoids, KNN, dendrograms, Principal Component Analysis and more modern methods such as t-SNE and UMAP, will allow us to solve these problems.
Predict the behavior of a variable over time, time series will be very useful.
Using algorithms such as Holt-Winters, ARIMA, GARCH or even neural networks will allow us to find solutions.
Solve problems related to losses, the risk that a certain type of investment maintains, maximum thresholds to cover the losses in some type of Operational Risk or another kind.
Additionally, find the actuarial present value, find the cost of premiums, annuities, pension amounts and others to know the sustainability of a pension fund or a company benefit plan.