5 Python Scripts for Time Series Analysis

A collection of standalone Python scripts covering the core tasks in time series analysis.

---

Scripts

Script	What it does
ts_resampler.py	Resample and aggregate irregular time series to a consistent frequency
ts_anomaly_detector.py	Detect anomalous data points using z-score, IQR, or rolling statistics
ts_decompose.py	Decompose a series into trend, seasonal, and residual components
ts_forecast.py	Fit a SARIMA model and generate a forward forecast with confidence intervals
ts_compare.py	Compare multiple series: correlation, lag analysis, summary statistics

---

Dependencies

Script	Packages
ts_resampler.py	pandas, openpyxl
ts_anomaly_detector.py	pandas, openpyxl, matplotlib
ts_decompose.py	pandas, openpyxl, statsmodels, matplotlib
ts_forecast.py	pandas, openpyxl, statsmodels, matplotlib
ts_compare.py	pandas, openpyxl, matplotlib, scipy

Install all at once

pip install pandas openpyxl matplotlib statsmodels scipy

Requires Python 3.10+.

---

Quick Start

1. ts_resampler.py

Align irregular time series data to a consistent frequency before any analysis.

# Resample to daily, fill gaps with forward-fill (default)
python ts_resampler.py --input sensor_data.csv --datetime-col "Timestamp" --freq D

# Resample to monthly, use interpolation for gaps
python ts_resampler.py --input readings.csv --datetime-col "ts" --freq MS --fill interpolate

# Per-column aggregation rules (sum sales, average temperature)
python ts_resampler.py --input data.csv --datetime-col "Date" --freq D --config agg_rules.json

Frequency strings: T/min (minutely), H (hourly), D (daily), W (weekly), MS (month start), QS (quarter start), AS (annual)

Example agg_rules.json:

{
    "Sales":       "sum",
    "Temperature": "mean",
    "WindSpeed":   "max",
    "EventCount":  "count"
}

Output: resampled.xlsx with Resampled, Summary, and Gap Report tabs.

---

2. ts_anomaly_detector.py

Flag data points that deviate from expected patterns using three detection approaches.

# Z-score method (default threshold: ±3 std deviations)
python ts_anomaly_detector.py --input data.csv --datetime-col "Date" --value-cols "Sales"

# IQR method across multiple columns
python ts_anomaly_detector.py --input data.csv --datetime-col "Date" \
    --value-cols "Temperature" "Pressure" --method iqr

# Rolling window method (better for trending series)
python ts_anomaly_detector.py --input data.csv --datetime-col "ts" \
    --value-cols "Revenue" --method rolling --rolling-win 30

# Run all three methods, save anomaly charts
python ts_anomaly_detector.py --input data.csv --datetime-col "Date" \
    --value-cols "Sales" --method all --plot

Methods:

• zscore — flags points beyond N standard deviations from the global mean
• iqr — flags points outside 1.5× (configurable) the interquartile range
• rolling — flags points deviating from a local rolling mean; handles trends better
• all — runs all three and flags any point caught by any method

Output: Annotated Excel file with _anomaly flag columns; red rows = anomalies.

---

3. ts_decompose.py

Separate a series into its underlying components for clearer analysis.

# Monthly data with annual seasonality
python ts_decompose.py --input sales.csv --datetime-col "Date" --value-col "Revenue" --period 12

# Daily data with weekly seasonality, multiplicative model
python ts_decompose.py --input traffic.csv --datetime-col "Date" --value-col "Visits" \
    --period 7 --model multiplicative

# Resample to month-start before decomposing, save chart
python ts_decompose.py --input raw.csv --datetime-col "ts" --value-col "Value" \
    --freq MS --period 12 --plot

--model options:

• additive — use when seasonal variation is roughly constant in magnitude
• multiplicative — use when seasonal variation scales with the trend level

Common --period values: 7 (daily data, weekly season), 12 (monthly data, annual season), 4 (quarterly data, annual season), 24 (hourly data, daily season)

Output: decomposed.xlsx with Components and Statistics tabs; optional 4-panel chart.

---

4. ts_forecast.py

Generate a forward forecast using SARIMA with validation metrics.

# Basic forecast, 12 periods ahead
python ts_forecast.py --input sales.csv --datetime-col "Date" --value-col "Revenue" --periods 12

# Auto-select best model parameters
python ts_forecast.py --input sales.csv --datetime-col "Date" --value-col "Revenue" \
    --periods 6 --auto-order

# Specify SARIMA order manually, save forecast chart
python ts_forecast.py --input data.csv --datetime-col "Date" --value-col "Sales" \
    --order 1 1 1 --seasonal-order 1 1 1 12 --periods 12 --plot

# Non-seasonal ARIMA (set seasonal period to 0)
python ts_forecast.py --input data.csv --datetime-col "Date" --value-col "Value" \
    --order 2 1 2 --seasonal-order 0 0 0 0 --periods 8

SARIMA parameters:

• --order p d q — AR order, differencing, MA order
• --seasonal-order P D Q s — seasonal AR, differencing, MA, and period
• --auto-order — grid search over common parameter combinations, picks lowest AIC
• --test-periods — number of hold-out periods used for MAE/RMSE/MAPE validation

Output: forecast.xlsx with Forecast, Forecast Only, and Model Metrics tabs; optional chart. Green rows = forecast values.

---

5. ts_compare.py

Analyse relationships between multiple time series.

# Compare three series, resample to monthly
python ts_compare.py --input data.csv --datetime-col "Date" \
    --value-cols "Sales" "Traffic" "Conversions" --freq MS

# Cross-correlation up to 6 lag periods
python ts_compare.py --input data.csv --datetime-col "Date" \
    --value-cols "LeadMetric" "LagMetric" --max-lag 6

# Compare all numeric columns, save dual-axis charts for top pairs
python ts_compare.py --input data.csv --datetime-col "Date" --plot

Output tabs:

• Aligned Series — all series resampled to a shared frequency
• Summary Stats — mean, std, min, max, median, trend slope and direction per series
• Pairwise Corr — Pearson and Spearman correlations for every pair, with p-values
• Corr Matrix — wide-format Pearson correlation matrix
• Lag Analysis — peak cross-correlation and lag for each pair (identifies leading/lagging relationships)

---