Random Forest Ad Spend Prediction Using Viewership Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Limited availability of public ad spend data makes it difficult for companies to accurately predict ad spend for specific networks at specific dates and times, hindering the calculation of return on ad spend (ROAS).
Innovation Solution
A computer program product using a database of media program data, including known ad spend and viewership data, trains a random forest model to predict ad spend by identifying correlated variables and applying median CPM encoding, normalization, and feature engineering to estimate ad spend for unknown media programs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If only a very small percentage of public ad spend data is used for training, then data availability is limited, but prediction accuracy can still be achieved through advanced modeling techniques
Solution Approach 1:
The system performs preliminary data preparation by collecting and preprocessing available ad spend data, viewership data, and program metadata before model training. This includes cleaning, normalizing, and structuring the limited public data to maximize its utility for training the random forest model, thereby achieving accurate predictions despite data scarcity
Solution Approach 2:
The system transforms the limited ad spend data by creating derived features and parameters such as cost per mille (CPM), viewership ratings, and program characteristics. These transformed parameters enhance the predictive power of the model by extracting more information from the same limited data set, improving prediction accuracy without requiring more raw data
2Measurement precision
If a random forest model with multiple variables is used to predict ad spend, then prediction accuracy improves, but model complexity increases
Solution Approach 1:
The random forest model is segmented into multiple individual decision trees, each trained on different subsets of the data with different feature combinations. This segmentation allows the complex prediction task to be divided into simpler, more manageable components while maintaining high prediction accuracy through the aggregation of multiple simpler models
Solution Approach 2:
The system uses a random forest model that employs partial action by considering only a subset of features for each decision tree during training, rather than using all available features in every tree. This approach reduces the complexity of individual trees while maintaining overall model accuracy, as the ensemble of multiple trees with different feature subsets captures the complex relationships in the data
3Reliability
If feature engineering and normalization techniques are applied to the data, then model performance improves, but data processing time and computational complexity increase
Solution Approach 1:
The system performs feature engineering, normalization, and data preprocessing as preliminary actions before model training. By preparing the data in advance—including creating derived features, normalizing scales, and encoding categorical variables—the model can focus on learning patterns rather than processing raw data, improving performance while managing processing time through efficient precomputation
Data Source
AI summary
A computer program product is provided for predicting ad spend for a specific media program aired or streamed on a specific network at a specific date and time using a database of media program data that includes known ad spend for a subset of media programs, and viewership data for each of the media programs, including total viewership and viewership ratings. Each of the media programs is identified by its respective network, and date and time of airing or streaming. A random forest model is trained to predict ad spend using variables that are identified as being correlated with ad spend. The trained random forest model is then used to predict ad spend for a specific media program that is aired or streamed on a specific network at a specific date and time, and which has an unknown ad spend.