
You will learn the transaction data and Python codes used in this course. Check the attached files in the resource of the section and download them!
Understand the customer lifetime value concept, how historical data and machine learning predict future purchases and active periods, and how to rank customers for targeted promotions.
Explore data preparation for clv modeling in Python: check data shape, handle missing values, filter by positive quantity, compute monetary spend, convert invoice dates, and run describe for statistics.
Generate clv features from transaction data using the lifetimes package and pandas. Build a dataset with customer id, end date, and t, focusing on frequency, age, recency, and monetary value.
Explore the CLV model's statistical background by linking Poisson, geometric, and gamma distributions to a likelihood framework that estimates churn probability and future spend from purchase frequency.
Partition the data by time into calibration and holdout periods, add validation columns, and fit the BGF model on pre-date data; then compare predictions to holdout purchases using calibration_purchases_holdout_purchases plot.
Fit the gamma-gamma model with the Python lifetime package to estimate each customer's average transaction value, then use the predicted value to inform the customer lifetime value calculation.
Explore how gradient boosting builds tree models to predict residuals and refine forecasts, using Python implementations like XGBoost and LightGBM for customer lifetime value modeling.
Apply xgboost and gradient boosting to build a customer lifetime value model in python, using features like recency, frequency, customer age, and external data, with train-test split and mse evaluation.
Partition the data into training and testing sets (70 percent training, 30 percent testing) with X1, X2 features and y target, and monitor performance with a watchlist during Lightgbm model training to build a customer lifetime value prediction.
If you wish to start the data analytics career or apply machine learning expertise into business, this is the right course you must choose!
Here I will provide a series of lectures on a practical marketing AI model -- 'Customer Life Value Model', or CLV model. The method is also sometimes called the 'repurchase modeling'. I would say what you will learn is a very useful AI forecasting model for marketing campaign and promotion. Because the CLV models I am teaching in this course are currently widely used in retail banking, insurance, and other sales-related industries. Why? Since it helps business owners select the most valuable customers to get their business better and better!
The value of my course is mainly reflected in the following aspects:
1. The CLV models can be quickly created because the process and features for building models are very concise and efficient.
2. One can utilize the model to predict the customer's purchase behavior or purchase preference for a specific merchandise in a given future time period.
3. The CLV model can be used to predict the probability of customers' repurchase behavior.
4. The CLV model can be used to analyze the activity and loyalty of different customers -- help you solve customer retention problems.
5. Based on the output of the CLV model, business owners can calculate and rank the customer lifetime value.
The objective of the course is to let you master how to effectively use the big data and AI algorithms for intelligent marketing. For example, if you can successfully predict who will buy the commodities in the next month based on historical transaction data, then you would effectively apply some market strategies into these customers, like by launching advertisements, applying recommender systems for ‘cross sales’ or ‘cross recommendation’ At the same time, The business owners will also realize from the model's prediction who are not interested in the goods or services they are providing, perhaps they can adopt some other marketing strategies or promotion to make these silent customers become more 'active' or 'waked up'.
In addition to the business value you can absorb from the course, I also teach you some practical statistical, machine learning and AI algorithm knowledge and skills, combined with the Python programming coding. This mainly covers:
1. Various statistical distribution functions such Geometric / Negative Binomial used in the CLV models and interpretations.
2. Lifetime package in Python to create BG/NBD CLV model.
3. Different analytical and graphics tools in Lifetime package including implementation methods and interpretation.
4. Data exploration, cleaning and feature generation for CLV models with Python programming.
5. Model feature selection, feature engineering, cross validation and performance tracking.
6. Lecture on how to apply the third party data into CLV modeling.
7. Introduction of gradient boost tree algorithms’ framework and implementation including Xgboost and Lightgbm algorithm into CLV modeling.