Innovative System for Identifying Competitors

In the fiercely competitive hospitality industry, understanding your competitors is crucial for strategic positioning and pricing decisions. This project introduces a cutting-edge system designed to analyze and identify hotel competitors based on sophisticated algorithms and critical data features.

Core Objectives and Methodology

The system’s primary objective is to accurately determine potential competitors of a given hotel by analyzing similarity scores. These scores are computed using a blend of algorithms and key features:

• Algorithms Utilized: Hidden Markov Model (HMM) and Dynamic Time Warping (DTW). These algorithms are adept at handling time series data and uncovering underlying patterns in pricing and geographical trends.
• Features Considered: The system focuses on two pivotal features – rate value and geographical location. These are instrumental in determining the competitive landscape of a hotel.
• Similarity Score: The essence of this system lies in its ability to quantify similarity, with scores ranging from 0 to 1. A score of 1 denotes identical behavior, while 0 indicates no correlation.

Purpose and Advantage

• Hidden Markov Model (HMM): Efficiently models time series data, capturing hidden state sequences more effectively than direct similarity measurements, especially the data variance is different.
• Dynamic Time Warping (DTW): Measures similarity between time series sequences.
• Combination Benefit: The use of HMM and DTW together surpasses traditional Correlation Measures, especially when dealing with time series data exhibiting similar observable values or patterns.

Input Data Structure

• RDL Table: Includes hotel ID, arrival and shop dates, room type, and rate.
• Location Table: Contains hotel ID, latitude, and longitude.

Output Data Structure

• Outputs include self and competitor hotel IDs, final similarity score, and competitor identification.

Algorithm Workflow

1. Data Cleaning and Processing: Involves selecting valid days, identifying nearby hotels, filling missing rate data, and standardizing room type rates.
2. Daily Component Computation: Calculates daily median rates and predicts hidden state sequences using HMM.
3. Pace Component Computation: Standardizes rates by DTA (days to arrival), convolves rate sequences, and calculates median DTW scores.
4. Similarity Score Transformation: Converts DTW distances into similarity scores, combining them to provide a relative measure of similarity.

Key Processes

• Competitor Selection: Chooses top 50 nearest competitors within 26 km and processes rate data for up to 64 days.
• Standardization and Convolution: Utilizes specific algorithms for data standardization and convolution, crucial for accurate similarity score computation.
• Total Weight Computation: Normalizes similarity scores, allowing for impact categorization (e.g., low, medium, high).

Final Output

• The algorithm outputs a nuanced similarity score, offering a sophisticated, data-driven perspective on hotel competition, crucial for strategic decision-making in the hospitality sector.