prophet-rb 0.2.1 → 0.2.5

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 6469e3c9b0f7d026678845025284f37f7c72f875e19a5f245d9bbc745d348a38
-  data.tar.gz: f3be6a4488cf07a9b2b36f2ce8dd20e9c42aa4297b36b7c56d0ac16bbd6a28f2
+  metadata.gz: ec194b65cb4bfd061bac7868cad435fea98df148fc7e129d8562c55ed3c91fcc
+  data.tar.gz: 45b9357d04a6c01a8d4ad6ce7888065b946e39cbfa743b650a41d5b12cf9989e
 SHA512:
-  metadata.gz: 6476dc943d7a23bf74e84a8fec345009b4587e1e514e9b8e22b32fdd02e44a0f60870eb918b5cb097cd47f1da3e9ad8481098f6800736eadc31fffd7db266f75
-  data.tar.gz: e831ae81325d98e12a3b53bf3e68f87fdc953ad8cba1f22daca5cd375c822a9d39e5f192a4c40eeaf5c62c852955fce3b300680538988a7449069fc3de25fefb
+  metadata.gz: cd67e017dac94994c960a42b59fb55f2c33560d9233bb2764581cef3154bc03bd75ba29283f8e51788fcde356f682e9fe487a5506ccb82ed10ab373b11774989
+  data.tar.gz: a07444bc2d9d6a29b6b09303b5d4cef60709a4a4be0a31d2a4526f79366696875b72a36923747b673f05b686c857a78be803f2475fbbc89ffb3dac4a1c856c07

data/CHANGELOG.md

@@ -1,3 +1,20 @@
+## 0.2.5 (2021-07-28)
+
+- Added `anomalies` method
+
+## 0.2.4 (2021-04-02)
+
+- Added support for flat growth
+
+## 0.2.3 (2020-10-14)
+
+- Added support for times to `forecast` method
+
+## 0.2.2 (2020-07-26)
+
+- Fixed error with constant series
+- Fixed error with no changepoints
+
 ## 0.2.1 (2020-07-15)
 
 - Added `forecast` method

data/LICENSE.txt

@@ -1,7 +1,7 @@
 MIT License
 
-Copyright (c) 2020 Andrew Kane
 Copyright (c) Facebook, Inc. and its affiliates.
+Copyright (c) 2020 Andrew Kane
 
 Permission is hereby granted, free of charge, to any person obtaining
 a copy of this software and associated documentation files (the

data/README.md

@@ -10,7 +10,7 @@ Supports:
 
 And gracefully handles missing data
 
-[![Build Status](https://travis-ci.org/ankane/prophet.svg?branch=master)](https://travis-ci.org/ankane/prophet) [![Build status](https://ci.appveyor.com/api/projects/status/8ahmsvvhum4ivnmv/branch/master?svg=true)](https://ci.appveyor.com/project/ankane/prophet/branch/master)
+[![Build Status](https://github.com/ankane/prophet/workflows/build/badge.svg?branch=master)](https://github.com/ankane/prophet/actions)
 
 ## Installation
 
@@ -48,6 +48,12 @@ series = User.group_by_day(:created_at).count
 Prophet.forecast(series)
 ```
 
+Detect anomalies in a time series
+
+```ruby
+Prophet.anomalies(series)
+```
+
 ## Advanced API
 
 Check out the [Prophet documentation](https://facebook.github.io/prophet/docs/quick_start.html) for a great explanation of all of the features. The advanced API follows the Python API and supports the same features. It uses [Rover](https://github.com/ankane/rover) for data frames.
@@ -118,7 +124,7 @@ Plot the forecast
 m.plot(forecast).savefig("forecast.png")
 ```
 
-![Forecast](https://blazer.dokkuapp.com/assets/prophet/forecast-a9d43195b8ad23703eda7bb8b52b8a758efb4699e2313f32d7bbdfaa2f4275f6.png)
+![Forecast](https://blazer.dokkuapp.com/assets/prophet/forecast-77cf453fda67d1b462c6c22aee3a02572203b71c4517fedecc1f438cd374a876.png)
 
 Plot components
 
@@ -126,7 +132,7 @@ Plot components
 m.plot_components(forecast).savefig("components.png")
 ```
 
-![Components](https://blazer.dokkuapp.com/assets/prophet/components-b9e31bfcf77e57bbd503c0bcff5e5544e66085b90709b06dd96c5f622a87d84f.png)
+![Components](https://blazer.dokkuapp.com/assets/prophet/components-2cdd260e23bc89824ecca25f6bfe394deb5821d60b7e0e551469c90d204acd67.png)
 
 ## Saturating Forecasts
 

data/lib/prophet.rb

@@ -21,19 +21,22 @@ module Prophet
     Forecaster.new(**kwargs)
   end
 
-  # to add time support in future, see
-  # https://github.com/ankane/prophet/commit/06e3562835cbcf06b8431f3a91fe2618d4703eb7
   def self.forecast(series, count: 10)
     raise ArgumentError, "Series must have at least 10 data points" if series.size < 10
 
+    # check type to determine output format
+    # check for before converting to time
     keys = series.keys
-    bad_key = keys.find { |k| !k.is_a?(Date) }
-    raise ArgumentError, "Expected Date, got #{bad_key.class.name}" if bad_key
+    dates = keys.all? { |k| k.is_a?(Date) }
+    time_zone = keys.first.time_zone if keys.first.respond_to?(:time_zone)
+    utc = keys.first.utc? if keys.first.respond_to?(:utc?)
+    times = keys.map(&:to_time)
 
-    week = keys.map { |k| k.wday }.uniq.size == 1
-    month = keys.all? { |k| k.day == 1 }
-    quarter = month && keys.all? { |k| k.month % 3 == 1 }
-    year = quarter && keys.all? { |k| k.month == 1 }
+    day = times.all? { |t| t.hour == 0 && t.min == 0 && t.sec == 0 && t.nsec == 0 }
+    week = day && times.map { |k| k.wday }.uniq.size == 1
+    month = day && times.all? { |k| k.day == 1 }
+    quarter = month && times.all? { |k| k.month % 3 == 1 }
+    year = quarter && times.all? { |k| k.month == 1 }
 
     freq =
       if year
@@ -44,10 +47,20 @@ module Prophet
         "MS"
       elsif week
         "W"
-      else
+      elsif day
         "D"
+      else
+        diff = Rover::Vector.new(times).sort.diff.to_numo[1..-1]
+        min_diff = diff.min.to_i
+
+        # could be another common divisor
+        # but keep it simple for now
+        raise "Unknown frequency" unless (diff % min_diff).eq(0).all?
+
+        "#{min_diff}S"
       end
 
+    # use series, not times, so dates are handled correctly
     df = Rover::DataFrame.new({"ds" => series.keys, "y" => series.values})
 
     m = Prophet.new
@@ -56,6 +69,28 @@ module Prophet
 
     future = m.make_future_dataframe(periods: count, include_history: false, freq: freq)
     forecast = m.predict(future)
-    forecast[["ds", "yhat"]].to_a.map { |v| [v["ds"].to_date, v["yhat"]] }.to_h
+    result = forecast[["ds", "yhat"]].to_a
+
+    # use the same format as input
+    if dates
+      result.each { |v| v["ds"] = v["ds"].to_date }
+    elsif time_zone
+      result.each { |v| v["ds"] = v["ds"].in_time_zone(time_zone) }
+    elsif utc
+      result.each { |v| v["ds"] = v["ds"].utc }
+    else
+      result.each { |v| v["ds"] = v["ds"].localtime }
+    end
+    result.map { |v| [v["ds"], v["yhat"]] }.to_h
+  end
+
+  def self.anomalies(series)
+    df = Rover::DataFrame.new(series.map { |k, v| {"ds" => k, "y" => v} })
+    m = Prophet.new(interval_width: 0.99)
+    m.logger.level = ::Logger::FATAL # no logging
+    m.fit(df)
+    forecast = m.predict(df)
+    # filter df["ds"] to ensure dates/times in same format as input
+    df["ds"][(df["y"] < forecast["yhat_lower"]) | (df["y"] > forecast["yhat_upper"])].to_a
   end
 end

data/lib/prophet/forecaster.rb

@@ -75,8 +75,8 @@ module Prophet
     end
 
     def validate_inputs
-      if !["linear", "logistic"].include?(@growth)
-        raise ArgumentError, "Parameter \"growth\" should be \"linear\" or \"logistic\"."
+      if !["linear", "logistic", "flat"].include?(@growth)
+        raise ArgumentError, "Parameter \"growth\" should be \"linear\", \"logistic\", or \"flat\"."
       end
       if @changepoint_range < 0 || @changepoint_range > 1
         raise ArgumentError, "Parameter \"changepoint_range\" must be in [0, 1]"
@@ -602,6 +602,12 @@ module Prophet
       [k, m]
     end
 
+    def flat_growth_init(df)
+      k = 0
+      m = df["y_scaled"].mean
+      [k, m]
+    end
+
     def fit(df, **kwargs)
       raise Error, "Prophet object can only be fit once" if @history
 
@@ -624,6 +630,8 @@ module Prophet
 
       set_changepoints
 
+      trend_indicator = {"linear" => 0, "logistic" => 1, "flat" => 2}
+
       dat = {
         "T" => history.shape[0],
         "K" => seasonal_features.shape[1],
@@ -634,7 +642,7 @@ module Prophet
         "X" => seasonal_features,
         "sigmas" => prior_scales,
         "tau" => @changepoint_prior_scale,
-        "trend_indicator" => @growth == "logistic" ? 1 : 0,
+        "trend_indicator" => trend_indicator[@growth],
         "s_a" => component_cols["additive_terms"],
         "s_m" => component_cols["multiplicative_terms"]
       }
@@ -642,6 +650,9 @@ module Prophet
       if @growth == "linear"
         dat["cap"] = Numo::DFloat.zeros(@history.shape[0])
         kinit = linear_growth_init(history)
+      elsif @growth == "flat"
+        dat["cap"] = Numo::DFloat.zeros(@history.shape[0])
+        kinit = flat_growth_init(history)
       else
         dat["cap"] = history["cap_scaled"]
         kinit = logistic_growth_init(history)
@@ -655,12 +666,12 @@ module Prophet
         "sigma_obs" => 1
       }
 
-      if history["y"].min == history["y"].max && @growth == "linear"
+      if history["y"].min == history["y"].max && (@growth == "linear" || @growth == "flat")
         # Nothing to fit.
         @params = stan_init
         @params["sigma_obs"] = 1e-9
-        @params.each do |par|
-          @params[par] = Numo::NArray.asarray(@params[par])
+        @params.each do |par, _|
+          @params[par] = Numo::NArray.asarray([@params[par]])
         end
       elsif @mcmc_samples > 0
         @params = @stan_backend.sampling(stan_init, dat, @mcmc_samples, **kwargs)
@@ -671,8 +682,10 @@ module Prophet
       # If no changepoints were requested, replace delta with 0s
       if @changepoints.size == 0
         # Fold delta into the base rate k
-        @params["k"] = @params["k"] + @params["delta"].reshape(-1)
-        @params["delta"] = Numo::DFloat.zeros(@params["delta"].shape).reshape(-1, 1)
+        # Numo doesn't support -1 with reshape
+        negative_one = @params["delta"].shape.inject(&:*)
+        @params["k"] = @params["k"] + @params["delta"].reshape(negative_one)
+        @params["delta"] = Numo::DFloat.zeros(@params["delta"].shape).reshape(negative_one, 1)
       end
 
       self
@@ -739,6 +752,11 @@ module Prophet
       cap.to_numo / (1 + Numo::NMath.exp(-k_t * (t - m_t)))
     end
 
+    def flat_trend(t, m)
+      m_t = m * t.new_ones
+      m_t
+    end
+
     def predict_trend(df)
       k = @params["k"].mean(nan: true)
       m = @params["m"].mean(nan: true)
@@ -747,9 +765,11 @@ module Prophet
       t = Numo::NArray.asarray(df["t"].to_a)
       if @growth == "linear"
         trend = piecewise_linear(t, deltas, k, m, @changepoints_t)
-      else
+      elsif @growth == "logistic"
         cap = df["cap_scaled"]
         trend = piecewise_logistic(t, cap, deltas, k, m, @changepoints_t)
+      elsif @growth == "flat"
+        trend = flat_trend(t, m)
       end
 
       trend * @y_scale + Numo::NArray.asarray(df["floor"].to_a)
@@ -885,9 +905,11 @@ module Prophet
 
       if @growth == "linear"
         trend = piecewise_linear(t, deltas, k, m, changepoint_ts)
-      else
+      elsif @growth == "logistic"
         cap = df["cap_scaled"]
         trend = piecewise_logistic(t, cap, deltas, k, m, changepoint_ts)
+      elsif @growth == "flat"
+        trend = flat_trend(t, m)
       end
 
       trend * @y_scale + Numo::NArray.asarray(df["floor"].to_a)
@@ -899,6 +921,9 @@ module Prophet
       # TODO add more freq
       # https://pandas.pydata.org/pandas-docs/stable/user_guide/timeseries.html#timeseries-offset-aliases
       case freq
+      when /\A\d+S\z/
+        secs = freq.to_i
+        dates = (periods + 1).times.map { |i| last_date + i * secs }
       when "H"
         hour = 3600
         dates = (periods + 1).times.map { |i| last_date + i * hour }

data/lib/prophet/version.rb

@@ -1,3 +1,3 @@
 module Prophet
-  VERSION = "0.2.1"
+  VERSION = "0.2.5"
 end

data/stan/unix/prophet.stan

@@ -73,6 +73,15 @@ functions {
   ) {
     return (k + A * delta) .* t + (m + A * (-t_change .* delta));
   }
+
+  // Flat trend function
+
+  vector flat_trend(
+    real m,
+    int T
+  ) {
+    return rep_vector(m, T);
+  }
 }
 
 data {
@@ -86,7 +95,7 @@ data {
   matrix[T,K] X;        // Regressors
   vector[K] sigmas;     // Scale on seasonality prior
   real<lower=0> tau;    // Scale on changepoints prior
-  int trend_indicator;  // 0 for linear, 1 for logistic
+  int trend_indicator;  // 0 for linear, 1 for logistic, 2 for flat
   vector[K] s_a;        // Indicator of additive features
   vector[K] s_m;        // Indicator of multiplicative features
 }
@@ -104,6 +113,17 @@ parameters {
   vector[K] beta;           // Regressor coefficients
 }
 
+transformed parameters {
+  vector[T] trend;
+  if (trend_indicator == 0) {
+    trend = linear_trend(k, m, delta, t, A, t_change);
+  } else if (trend_indicator == 1) {
+    trend = logistic_trend(k, m, delta, t, cap, A, t_change, S);
+  } else if (trend_indicator == 2) {
+    trend = flat_trend(m, T);
+  }
+}
+
 model {
   //priors
   k ~ normal(0, 5);
@@ -113,19 +133,10 @@ model {
   beta ~ normal(0, sigmas);
 
   // Likelihood
-  if (trend_indicator == 0) {
-    y ~ normal(
-      linear_trend(k, m, delta, t, A, t_change)
-      .* (1 + X * (beta .* s_m))
-      + X * (beta .* s_a),
-      sigma_obs
-    );
-  } else if (trend_indicator == 1) {
-    y ~ normal(
-      logistic_trend(k, m, delta, t, cap, A, t_change, S)
-      .* (1 + X * (beta .* s_m))
-      + X * (beta .* s_a),
-      sigma_obs
-    );
-  }
+  y ~ normal(
+  trend
+  .* (1 + X * (beta .* s_m))
+  + X * (beta .* s_a),
+  sigma_obs
+  );
 }

data/stan/win/prophet.stan

@@ -47,7 +47,7 @@ functions {
     }
     return gamma;
   }
-  
+
   real[] logistic_trend(
     real k,
     real m,
@@ -94,6 +94,17 @@ functions {
     }
     return Y;
   }
+
+   // Flat trend function
+
+    real[] flat_trend(
+    real m,
+    int T
+  ) {
+    return rep_array(m, T);
+  }
+
+
 }
 
 data {
@@ -107,7 +118,7 @@ data {
   real X[T,K];         // Regressors
   vector[K] sigmas;     // Scale on seasonality prior
   real<lower=0> tau;    // Scale on changepoints prior
-  int trend_indicator;  // 0 for linear, 1 for logistic
+  int trend_indicator;  // 0 for linear, 1 for logistic, 2 for flat
   real s_a[K];          // Indicator of additive features
   real s_m[K];          // Indicator of multiplicative features
 }
@@ -135,6 +146,8 @@ transformed parameters {
     trend = linear_trend(k, m, delta, t, A, t_change, S, T);
   } else if (trend_indicator == 1) {
     trend = logistic_trend(k, m, delta, t, cap, A, t_change, S, T);
+  } else if (trend_indicator == 2){
+    trend = flat_trend(m, T);
   }
 
   for (i in 1:K) {

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: prophet-rb
 version: !ruby/object:Gem::Version
-  version: 0.2.1
+  version: 0.2.5
 platform: ruby
 authors:
 - Andrew Kane
-autorequire: 
+autorequire:
 bindir: bin
 cert_chain: []
-date: 2020-07-15 00:00:00.000000000 Z
+date: 2021-07-29 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: cmdstan
@@ -52,78 +52,8 @@ dependencies:
     - - ">="
       - !ruby/object:Gem::Version
         version: '0'
-- !ruby/object:Gem::Dependency
-  name: bundler
-  requirement: !ruby/object:Gem::Requirement
-    requirements:
-    - - ">="
-      - !ruby/object:Gem::Version
-        version: '0'
-  type: :development
-  prerelease: false
-  version_requirements: !ruby/object:Gem::Requirement
-    requirements:
-    - - ">="
-      - !ruby/object:Gem::Version
-        version: '0'
-- !ruby/object:Gem::Dependency
-  name: rake
-  requirement: !ruby/object:Gem::Requirement
-    requirements:
-    - - ">="
-      - !ruby/object:Gem::Version
-        version: '0'
-  type: :development
-  prerelease: false
-  version_requirements: !ruby/object:Gem::Requirement
-    requirements:
-    - - ">="
-      - !ruby/object:Gem::Version
-        version: '0'
-- !ruby/object:Gem::Dependency
-  name: minitest
-  requirement: !ruby/object:Gem::Requirement
-    requirements:
-    - - ">="
-      - !ruby/object:Gem::Version
-        version: '5'
-  type: :development
-  prerelease: false
-  version_requirements: !ruby/object:Gem::Requirement
-    requirements:
-    - - ">="
-      - !ruby/object:Gem::Version
-        version: '5'
-- !ruby/object:Gem::Dependency
-  name: daru
-  requirement: !ruby/object:Gem::Requirement
-    requirements:
-    - - ">="
-      - !ruby/object:Gem::Version
-        version: '0'
-  type: :development
-  prerelease: false
-  version_requirements: !ruby/object:Gem::Requirement
-    requirements:
-    - - ">="
-      - !ruby/object:Gem::Version
-        version: '0'
-- !ruby/object:Gem::Dependency
-  name: matplotlib
-  requirement: !ruby/object:Gem::Requirement
-    requirements:
-    - - ">="
-      - !ruby/object:Gem::Version
-        version: '0'
-  type: :development
-  prerelease: false
-  version_requirements: !ruby/object:Gem::Requirement
-    requirements:
-    - - ">="
-      - !ruby/object:Gem::Version
-        version: '0'
-description: 
-email: andrew@chartkick.com
+description:
+email: andrew@ankane.org
 executables: []
 extensions:
 - ext/prophet/extconf.rb
@@ -148,7 +78,7 @@ homepage: https://github.com/ankane/prophet
 licenses:
 - MIT
 metadata: {}
-post_install_message: 
+post_install_message:
 rdoc_options: []
 require_paths:
 - lib
@@ -163,8 +93,8 @@ required_rubygems_version: !ruby/object:Gem::Requirement
     - !ruby/object:Gem::Version
       version: '0'
 requirements: []
-rubygems_version: 3.1.2
-signing_key: 
+rubygems_version: 3.2.22
+signing_key:
 specification_version: 4
 summary: Time series forecasting for Ruby
 test_files: []